Files
claude/.claude/tasks/TODO.md
T

90 KiB
Raw Blame History

TODO

Open (reconciled 2026-11-12 before 3.0.0, /reconcile on the BDR-125 phases; full pass 2026-10-23)

Done history before 2026-10-08 removed 2026-10-23 (git keeps it: git log -p -- .claude/tasks/TODO.md); decisions live in .claude/memory/.

BDR-125 close-out run (user 2026-11-12): no phase 0/0b/4a/1/5/2/4b/4c/4d line left in TODO or BUGS-FOUND

Decisions at start (asked once): faunosteo memory v2 on chore/memory-v2 off develop in a fresh worktree (split already on develop 7a75974, 11 rules); settings.json npm -g deny lines removed on the block E branch, classifier test = npm install -g --dry-run cowsay; 3.0.0 cut right after block A (/release-candidate), later blocks land under the next Unreleased; LRN-270 bisect = user-run interactive sessions, one --settings file per pack. Merge prompts (gitflow finish ask) = the human signal per block once gates are green.

  • A merge chore/memory-feedback-gate1 (b7505d1 + routing confirmed row) into develop
  • A' /release-candidate 3.0.0 (scope as ticked 2026-11-11)
  • B chore/phase4-proofs: known-error replay (memory: known BLK-xxx + tool record), live redaction (fake secret → ledger REDACTED), digest after a real /compact, /context full+dev delta vs the 2026-11-09 base; each = command + output + journal line; a failure → /bugfix
  • C /feat gate 1: idle trigger tightened (≥ 8 words AND a real idle gap), recall cap 2, body indexed for BM25 with p90 before/after (> 3 s → back to title+lesson, "accepted"), 4d.4 built (/prune-memory step D: noisy rewrite + BDR-100 promotion)
  • D /feat D1b: digest falls back on the first unchecked items when ### Actionable is absent; init-project + onboard templates emit the heading; local repos' TODO migrated on chore/* (one per repo, merge on ask); Session-start step 1 "Read TODO.md" dropped from the doctrine, citers census
  • E phase 1 + 2 decisions: design hook DELETED (web-building.md loads by path); "Skill routing" lines → skills' description:; npm -g deny lines out + dry-run test; rtk cat|head|tail *.env* forms benched, every form kept without proof of obsolescence, T42 list updated; BLK-022 buried (phase 2 covers it): guard-bash.test.sh SKIP spec deleted, BLK-022 → superseded, TODO "Blocked" block removed
  • F faunosteo chore/memory-v2: memory.sh upgrade, /context before/after, /prune-memory there for BDR-080 (noisy) + LRN-080 (promotion); never merged by me
  • G BLK-044: trace oom_score_adj origin (unit, cc.sh), OOMScoreAdjust on the cc scope, proof via /proc/<pid>/oom_score_adj
  • H residuals in order: phase 0 (project-rules.sh emit/extract/..), 4a (BLK-038, MIGRATION.md:209), 5 (fragment key allowlist), 2 LOWs (awk, mkdir -m 700, cat x.key, doctor.sh:255), 4d follow-ups (transcripts per (root, session, agent), image prompts, threshold boundary tests, tsc, l:<root> one global key), BUGS-FOUND unfixed since 11-09, LRN-270 bisect (user-run), doctor skillOverrides residue; a residual NOT done = justified in the TODO line removal
  • I model-router live checks: W4-B AC9 + parentModel done from transcripts/logs; Fable-out-of-credits + StopFailure closed "not reproducible, reopen on incident"
  • J /reconcile, CHANGELOG (gate 1 at 57 % delivered + each block), 0 BDR-125 line left, EVAL of the run, /close

Context mod queue (BDR-125): one /ship-feature per phase, order (user 2026-10-25): 0 → 4a → 1 → 5 → 2 → 3 (conditional) → 4b-4d

Re-challenged 2026-10-25 on Fable (4 lenses + native-features guide, synthesis in .claude/tasks/plans/2026-10-25-context-mod-queue-challenge.md § RE-CHALLENGE; user picks: native ask merge gate, phase 3 conditional, rest in block). The plan halves: native carriers and existing files first, a mod only where a measured gain has no native lever. Shared rules for every phase:

  • Baseline first: claude -p "/context" at repo root (2026-10-23: 30.2k always-on = memory files 11.6k, skills 10k, agents 2.6k, system 2.5k, MCP instr 2.2k, tools 1.1k). Same measure after; the delta is a phase criterion.

  • Proof budget: a moved path rule gets a STATIC check (frontmatter parses, each glob matches ≥1 git ls-files path per project) + one session logged by the InstructionsLoaded hook (load_reason: path_glob_match). Model canaries (bench .claude/tasks/plans/2026-10-24-context-mod-probe/) only for hook and mod levers; the path lever is proved three times already.

  • Subtract-only: source text stays in CLAUDE.global.md / rules / CLAUDE.md. A doctrine a mod replaces is jit-marked, never deleted (mod off = full text). Guards fail closed, injections fail open. Status model everywhere: ONE parser, "last status-bearing line wins", the entry's own header [tag] read first.

  • Never: forced continuation / Stop block (BDR-083, BDR-087), per-turn nag (BDR-037), forced skill invocation, hiding a dispatchable agent via agent.offer (breaks dispatch, proved), a hard_deny that names a path, file or source line (the auto classifier would refuse all mod maintenance): hard_deny names off switches only ("enabled": false, enabledPlugins, disableAllHooks) and "never edit permissions.ask/deny to pass a gate".

  • Cleanup is part of each phase: delete what the phase makes obsolete (scripts, hooks, tests, doctrine lines, Makefile targets, doctor checks), grep every citer and patch in the same branch (doctrine-citers census in make test).

  • Phase 4b-4d, memory v2 capture, read, feedback (shrunk 2026-10-25 to the measured channels)

    • Measured recall (5 real FR prompts, 22 relevant entries): BM25 full text 9/22, haiku rerank of lexical top-30 14/22 (~8.5k chars). Error 4-gram match on real errors: BLK-030/031/034 exact, no false positive; 5 of 7 real errors of 2026-10-23 had NO entry: capture is the weak link.
    • 4b capture (mod). Failure ledger: tool results with isError, deny or non-zero exit carrying a distinctive error line → normalized + verbatim + command, REDACTED at capture (rtk-rewrite redaction + gitleaks patterns), bounded ring per session in $.store (4 MiB cap), matched against the registry error index; unmatched = new failure modes. Milestones: feat/fix commit, gitflow finish, a suite red then green. memory_pending tool returns unmatched failures + milestones; /capitalize, /close and every CAPITALIZE step call it first; one context note at the next prompt after a milestone, never per turn, never at Stop. "Proactive capitalization" paragraph KEPT. Writing always = model draft + user approval + lib/memory.sh.
      • Built 2026-11-04 on feature/memory-capture (f1d4f9d..HEAD): mod mods/memory, kit 100+ cases, consumers wired, noise measured (NOISE.md); merged 2026-11-04 into develop (dfadf8a): GATE 0 MET, verifier criteria 1-15 met (9 proven live), security PASS ×4, BDR-137
      • shortcut (done in 4c, Task 1: the lexical reader is gone, statuses come from memory.sh index): mods/memory/hooks/register.ts isSuperseded is a lexical reader, not memory.sh's status parser; upgrade = share that parser when 4c needs it (final review M11)
    • 4c read side (built 2026-11-05 on feature/memory-recall, merged 2026-11-06 ddb04f4). Two channels only: intent recall at prompt.submit for classifier labels design/analysis or the first coding prompt after idle (BM25 top-30 over title + lesson + body → haiku keeps ≤3, bounded 3 s, fail open) and tool-error 4-gram match (PostToolUse; a classic hook is acceptable if the mod ledger is not needed). Session digest (open BLK via the status parser, actionable TODO, last journal line) in hooks/session-start.sh on source: startup|compact|clear (native matchers; the mod's session.compact only skips); a SubagentStart twin if sub-agents need it. Injection = ID + title + lesson line; superseded never injected; budget ~1.5k chars per turn; once per context. Deferred: Read path triggers, Bash command triggers, "critical → deny once", Tier 1, cross-repo scope=global index, memory_search tool.
    • 4c follow-up from the 4b live proof: memory_pending lists earlier-session milestones marked (earlier session) (done 78eff1c 2026-11-05, mods/memory/hooks/register.ts milestoneLines).
    • 4d feedback (4d.1-4d.3 built 2026-11-08 on feature/memory-feedback, BDR-140, merged 2026-11-08 dcb593a). Mod logs cands+ms; bun reader lib/memory-feedback/; /memory-feedback skill; banner reminder; two gates (prompt 20 or 35 d, tool 10 or 35 d +1 extension). Plan r3 docs/superpowers/plans/2026-11-06-memory-feedback.md (purged at finish; git keeps it), contract .claude/tasks/contracts/2026-11-06-memory-feedback-4d-1825.md. 4d.4 (/prune-memory step D, BDR-100 promotion) stays conditional: built only when a gate reports a noisy entry or a promotion candidate.
    • 4d follow-ups (final review + task reviews 2026-11-08, non-blocking): transcript cache holds every parsed file per run (~1 GB at steady state: process per (root, session, agent) and drop); image prompts (array content) become lost (counted); no boundary tests for the threshold actions (3rd pick, late, tool, miss, noisy/promotion, idle tie-break), eviction note, torn count, toolItem; replyAfter hides seen.add in a filter; no tsc type check on lib/memory-feedback; kit test pinning the exact field set of a prompt log record (IR11 is toMatchObject since 4d.1, so a future text field would pass); labels.ts minors (fromLine checks column count only, readState accepts non-object JSON, .bak keeps one state, misleading error on an entry label with a cause).
    • shortcut: the injection log lives under l:<root> store keys (about 28 KiB each, 200 records and 64 KiB at most); sweepLogs counts them with the p: keys and evicts the oldest log key (by its newest record, never the current root's) once the total nears the engine's 4 MiB store cap; upgrade = one global log key or a smaller per-root cap (mods/memory/hooks/register.ts).
    • residual: the pre-existing memory.sh check call in hooks/session-start.sh is not time-bounded (the digest's memory.sh index blocker call is); upgrade = wrap it in timeout 2. Done 2026-11-06 d93f06c (SD16).
    • BLK-043 hotfix merged 2026-11-06 (bd5e661 main, 7e47bea develop, back-merged into feature/memory-recall 3564e3e, worktree removed).
    • BLK-043 Part B: suite-wide claude PATH stub + final empty-log check in lib/tests/session-digest.test.sh (SD0/SD15, done 2026-11-06 via /hotfix; security PASS, MEDIUM folded into the shortcut below).
    • shortcut (BLK-043 gate 2 LOWs): lib/tests/detect-plan.test.sh unchecked mktemp -d (empty WORK → real claude probe), bash -c "source '$LIB'" interpolation, no GIT_CEILING_DIRECTORIES on hook runs; and the memory, graphify-gate, push-guard suites run the hook with no claude stub. Upgrade = || exit 1 + command -v claude assert, positional args, ceiling dir, shared stub helper. Same gap in lib/tests/session-digest.test.sh SD0 (gate MEDIUM 2026-11-06): claude probe runs before any check that the stub is first on PATH (stub write failure or noexec TMPDIR → real CLI), a red SD0 does not stop the suite; LOW: log path interpolated into the generated stub unquoted. Done 2026-11-06 dc5f127 + 7ca6b18 (shared lib/tests/claude-stub.sh, 6 suites, BLK-045).
    • BLK-044: trace origin of oom_score_adj=-1000 on claude processes; set OOMScoreAdjust on the cc scope so a runaway claude dies first.
    • safety-residuals gate LOWs (2026-11-06, bugfix/safety-residuals): rtk read -n server.key / rtk read id_rsa at repo root pass (.key/.pem/id_* rtk denies all need */; pre-existing, bench before adding); session-start memory.sh check and digest unbounded when timeout is absent (stock macOS); pre-commit allowlist fails open if git diff --cached errors (pre-existing, Gitea backstop); lib/memory.sh _mem_path_safe line > 80 chars; claude-stub.sh interpolates the mktemp log path into the generated stub (test-only).
    • Proof: recall precision on ≥20 real prompts and ≥10 real errors, target ≥ the 14/22 baseline; digest after a real compaction; redaction test on the ledger; /context delta; captured-failure demo: an error from session N recalled in session N+1.
    • Gate 1 result (2026-11-12, EVAL-058): prompt precision 57 % (n=75), under the 64 % target; fresh 78 %, idle 44 %, keyword 55 %; tool gate 0 records.
    • 4d gate 1 (2026-11-12, EVAL-058): idle trigger at 44 % precision: tighten (skip prompts under N words, or only after a real idle gap) or drop; decide before the next gate
    • 4d gate 1: cap recall at 2 entries (3rd pick 35 %)
    • 4d gate 1: tool channel logged 0 records in 4 days: check the PostToolUse error match actually fires (live error in a session, then memory-feedback)
    • 4d gate 1: misses 4/7 no-overlap: BM25 over title+lesson cannot reach them; option = index body too, or accept

BDR-125 residuals (after 3.0.0)

Open children of the shipped phases (History 2026-11-12), text verbatim.

  • Phase 0 / 0b leftovers (extract-hardening 14a93e3, lib/project-rules.sh):
    • emit: reuse _rules_out_path_ok (today -e only, a dangling link at the target is followed) and check the mkdir -p and > "$2" rc (today "emitted" rc 0 on a failed write).
    • extract rewrite failure mid-cat (ENOSPC/EIO after the truncation of $src): keep $tmp and name it on stderr instead of rm -f (today "$src not rewritten" can be false, the two copies gone; onboarder S1 rollback covers the split path, not direct lib use).
    • shortcut _rules_out_path_ok: the .. check is lexical, a symlinked parent component (.claude or .claude/rules pointing elsewhere) is not resolved; upgrade = cd -P on the parent dir, compare with $PWD (no portable realpath -m).
  • Phase 4a leftovers:
    • faunosteo: NOT upgraded (V1-2 active in another session's worktree; user: no write to develop or the V1-2 worktree). Patch prepared off-repo from V1-2 587cbb2 (rc 0, 520 lines, open BLK-008/009 unchanged); at integration re-run memory.sh -C <faunosteo> upgrade on the branch the user names rather than applying the stale patch. Stale clone /home/bchanot/faunosteo (main, 2026-07-01): left alone, user to decide.
    • BLK-038 residuals (gates 2026-10-30, non-blocking): v1 line echoed on stderr unfiltered (CWE-150 LOW, terminal control chars from a foreign registry; same as existing "upgrade lost" path); no fixture asserts the "wrote upgrade status" / "dated status line kept" messages; the real-data replay oracle lives in the session scratchpad (up4a/realdata.sh), not re-runnable: port to a make target over fixture copies if the upgrade is touched again; memory.sh journal leaves two blank lines before a new date heading on files ending with a blank line (commit-changer note).
    • MIGRATION.md:209 shows memory.sh -C <project> add … --body-file … with no absolute-path note (same trap as § WRITE, parked by the docs-half challenge 2026-11-01): next doc pass.
  • Phase 1 leftovers:
    • Deferred (gate 2026-10-29): "Skill routing" lines → our skills' description: (phase 3 bullet said phase 1 cleanup).
      • Design hook false positives: the pointer is almost never followed, and it fires on non-design prompts (fired on the 2026-10-30 phase 4a memory-work prompt). Decide: tighten the trigger, or drop the hook now that web-building.md loads by path. User call, nothing coded.
  • Phase 5 leftovers:
    • Follow-ups: lib/pack-census.sh:67 grep -qx → grep -qxF; fragment top-level key allowlist in lib/pack-census-fragments.jq (today only enabledPlugins keys are compared). Done 2026-11-02: MIGRATION step 2 $root re-derivation + autostash-conflict note (0463583, 9cd9591).
  • Phase 2 leftovers:
    • Phase 2 residuals (stated 2026-11-02): Bash write forms are outside the routing guard (the harness auto-mode "prefer Bash" path); a register.ts reload resets the skill latch; .claude/worktrees/ writes are code; code-limits state dirs stay in TMPDIR; not caught by a static rule: … | bash -s, a sourced gitflow_finish, git -C <dir> merge (soft_deny prose only).
    • Phase 2 security gate follow-ups (2026-11-02, non-blocking): MEDIUM rtk read -n .env / rtk read a.txt .env typed directly pass every static deny (rc 3 → passthrough, no rule) → bench Bash(rtk read * .env*), Bash(rtk read * */.env*) + */rtk twins and teach phase2.jq the * <path> shape; LOW code-limits state dir (mkdir -m 700, owner check); LOW awk evasions (system (, awk -f, gawk/mawk); LOW pre-existing cat x.key (no dir prefix) parity gap; doctor.sh:255 jq message should name code-limits and push-guard. MEDIUM rtk read part done 2026-11-06 97c6afc (2 rules, bench R3-R7, LRN-266); the LOWs stay open.
    • rtk deny forms (user 2026-11-12, split from the line above): the 6 rtk cat|head|tail *.env* deny forms (Bash(rtk cat *.env*), Bash(*/rtk cat *.env*), head, tail) said obsolete; removing them needs a verified reason per form and reviewed rows in the push-guard T42 justified list, as its own run.
  • Context measure leftovers (2026-11-09, LRN-270):
    • Open: dev System tools 11.3k vs full 7.7k (interactive) comes from some pack being off, not MCP; unattributable via /context. Bisect by toggling one pack at a time interactively if it matters.
    • Follow-up candidate (accepted risk): doctor residue check covers enabledPlugins only; a UI write into user-scope skillOverrides would dirty tracked settings.json unseen.

Actionable

  • BDR-125 obsolescence pass (2026-11-09, /code-clean merged db5bbe5; report .claude/tasks/plans/2026-11-09-bdr125-obsolescence-clean-1650.md): D1-D6, G1-G5 and the D3 add-ons done and merged by 2026-11-12; still open below: D1b and the G2 measured passive-cost feature.

    • D1 doctrine (done c69ce9c + docs a37df2f, feature/doctrine-rules-graphify; TODO.md stays read at start, see D1b): CLAUDE.global.md:52-54 "Session start 1. Read .claude/memory/" + skills/capitalize/SKILL.md:375, while the 4c digest delivers open BLK + TODO + last journal at start (4a meant to fix the "load every session" premise)
    • D1b (user 2026-11-12, after D1 challenge): every project TODO readable by the session digest. _digest_todo (hooks/session-start.sh:329-333) reads only a ### Actionable heading, which only this repo has; skills/init-project seeds # TODO\n\n## <date>, skills/onboard STEP 9 writes ## P0 — Critique…. Options to choose at planning: templates/init-project/onboard emit the heading + migrate existing TODOs (game, faunosteo, zenquality…), and/or digest falls back to the first unchecked - [ ] items. Until then D1 keeps TODO.md read at session start.
    • D2 visible (done 80f56ef, feature/banner-drop-plugin-dev): banner "OFF : plugin-dev" (plugin dropped 2026-07-02): lib/detect-plugins.sh:43-47, hooks/session-start.sh:106,111,180,210, agents/plugin-advisor.md:59,181,199,242,250-255
    • D3 visible (done 9b12b12, BDR-142; follow-ups in BUGS-FOUND) (= docs backlog P34): packs vocabulary in agents/plugin-advisor.md (:99 PROFILE template in the old profile.sh current format, :126-129 rollback, gstack ON/OFF rules, context7 ON) and USAGE.md:181,232,233,1090,1092,1095
    • D4 skill behavior (done 99f10b2 + docs bd0687d, feature/prune-memory-status-reader; one classifier reconcile_status_classify, unreadable = info-only count; follow-ups in BUGS-FOUND): skills/prune-memory STEP 1.A/3 detection (per-entry frontmatter, raw status: lines) → read statuses via memory.sh index (v1 + v2); keep example IDs 9xx (RED-7)
    • D5 user rules (done c69ce9c; IDOR + RLS kept, service-key line dropped): rules/web-security.md vs rules/secure-by-default.md: strict duplicates (web-security :11, :19, "Login rate limiting," in :20) and password wording conflict ("bcrypt/argon2" vs "argon2id (bcrypt or scrypt otherwise)"); cookie flags, upload type/size, forced HTTPS, security headers are unique to web-security
    • G1+G4 fixed (bugfix/banner-effective-profile, merged a655b06); G3 fixed (merged 9f9821e); G2 hand figures removed (bugfix/passive-cost-source, plan .claude/tasks/plans/2026-11-10-passive-cost-source-1330.md r3); G5 fixed (bugfix/handover-memory-status, 93a8e19, plan .claude/tasks/plans/2026-11-10-handover-memory-status-1600.md r3); follow-ups G6-G9 + 2 LOW quoting in BUGS-FOUND
    • feature (G2 split, user 2026-11-10): measured passive-cost API, one source for doctor / /status / plugin-advisor (probe COSTS field). Spec input = the two challenge rounds in the G2 plan: unit = pack or ON plugin (skills + agents + commands descriptions; pr-review-toolkit ~1550t, example-skills ~2350t), doctor counts OFF packs (higgsfield, extras), detect_plugin_dev globs the wrong cache level and ignores enabled state, ? for not-installed or empty units, rc-0 sentinels (costs-unavailable/unknown) for doctor's set -euo pipefail, profile.sh failure → unknown, python3 -I (probe runs in the user's project), installPath from installed_plugins.json. Doctor keeps its measured ui-ux 670 until then.
    • D3 add-ons (done 9b12b12) from G2 (plugin-advisor.md): plan-budget WARN :317 (unreachable: plugin-gate never sends PLAN, detect_plan gone), :42 "absolute COST ESTIMATE still reported", :105 (budget: ~<N>t | n/a), :222 "passive cost estimate = highest-cost package profile", "0t passive" claims for impeccable / design-motion-principles (:226 :227 :268 :269); USAGE.md :248 "Budget Pro (~11k tokens/5h)", :494 (~400t), :923 (~200t); session-banner.test.sh SB10 gone() passes on a missing file (add an existence check)
    • D6 project rule (done c69ce9c; lead paragraph fixed too, follow-ups in BUGS-FOUND): .claude/rules/graphify-vendored.md says graphify install --platform claude "touches nothing else", but graphifyy 0.9.67 appends a # graphify block to ~/.claude/CLAUDE.md (= CLAUDE.global.md); graphify claude install writes into the caller's cwd, so the install-plugins.sh trap protects $REPO only when run from $REPO
  • parked (code-clean 2026-11-09): hooks/rtk-rewrite.sh:84 no-op [ "$CMD" = "$REWRITTEN" ] && REWRITTEN="$CMD": remove with the next real rtk-hook edit, re-pin hooks/.rtk-hook.sha256 + rtk-rewrite.test.sh T13 in the same step

  • deferred (BDR-123): remap inline-writing best-tier skills (design Build stack, writing-skills, deprecation-and-migration, /tour single-project, /audit-delta 3d, /harden STEP 3); stale citers skills/feat/SKILL.md:6, skills/code-clean/SKILL.md:36; + impeccable row in routing.json (design Build stack has none; frontend-design has reflect; census test pins the table)

  • parked (GATE 2 LOW, 5a98c2c): probe table names not checked with IDENT.fullmatch per table (a / walks a path on the approved Supabase host); printable lets Unicode bidi/format controls (U+202E, U+2066…) through to the tty

  • profile name charset check (done by phase 5: NAME_RE='^[a-z0-9][a-z0-9-]*$' lib/profile.sh:63 on every profile and session name; the old read_profile() is gone)

  • P22 make new-skill scaffold: agent stub lacks effort: pin, skill stub lacks entry level, body loads .claude/agents/$(name).md instead of $HOME/.claude/agents/<name>.md → /hotfix; also: no routing.json row reminder

  • lib/tests/design-tool-gate.test.sh hermetic CLI_ABSENT case (done by the phase 5 rewrite: suite green 2026-11-02 with ~/.local/bin/21st installed, make test all suites green on this machine)

  • post-run-D residuals (gates, 2026-10-07): settings.json soft_deny names only gitflow.autopush false → add "or an unparseable value" (restriction only, settings run); capitalize STEP 6: manual mode with ahead = 0 (user pushed by hand between merge and read) matches no closing line, and the 5C ahead = 0 bullet still says "(auto-push mode did it)"; release-executor:86 line > 80 cols; push-guard header line 4 > 80 cols (pre-existing); unpushed-guard Stop silence now also covers a missing lib (SessionStart names it) — accepted

  • gitflow LOW leftovers (run B, rest done in run D 472cccb): --end-of-options/-- on refname args and printf %q in copy-paste hints; gitflow_delete: check _gitflow_checkout_containing_base rc (lib/gitflow.sh:270) before --unset-upstream

  • docker host-root mount spellings uncovered by deny (security gate 2026-10-23, pre-existing): --volume /:/x, -v=/:/x, --mount type=bind,source=/,...; prove each with a harmless analogue (--settings one-rule file) before adding. Not folded by phase 2 (2026-11-02): needs its own bench rows.

  • 6 dead deny entries (done: test 79f1a3f, settings 12c3534, merged 29cb080) (Bash(docker run * -v /:*), Bash(git push * :*), 4 Write(...)) kept by the push-guard T41/T42 non-weakening lock; drop them with a justified-removal list in T42 at the next release (working forms added 2026-10-23)

  • after 3.0.0: the next version bump must empty the push-guard removed-ok list (lib/tests/push-guard.test.sh:330 removed_ok_until=3.0.0), or T42e goes red.

  • after 3.0.0: /bugfix the 2 D3 MEDIUM gate findings (plugin-gate accepts pack extras, 21st-ai credit-metered; REQUEST keywords alone trigger embedded → minimal), both in BUGS-FOUND (user 2026-11-12: after the cut)

Release 3.0.0

Scope (user 2026-11-11) [gated]: D1-D6 + the 6 dead deny entries + MIGRATION/CHANGELOG land before the cut. Decisions: D1 digest + on-demand read (capitalize:375 aligned); D2 remove plugin-dev everywhere (OFF row disappears when empty); D3 advisor + USAGE rewritten to packs vocabulary, no budget/token claims, G2 add-ons included, SB10 fails on a missing file; D4 /prune-memory reads statuses through memory.sh (v1+v2), example IDs 9xx; D5 web-security keeps only its unique rules, points to secure-by-default, password = argon2id; D6 rule text names the CLAUDE.global.md append + install-plugins.sh runs the graphify step from $REPO. Runs: /feat D2+D3, /feat D4, /feat D1+D5+D6, /feat dead denies, docs, /release-candidate.

  • MIGRATION.md "Upgrading to 3.0.0" (done ed67c52, chore/docs-3.0.0) (CHANGELOG Unreleased says breaking: typed /effort-* removed; only the 2.0.0 section exists)
  • routing.json confirmed (15 agents, 11 phases, 6 skills) ships to every clone: kept as is (user 2026-11-11, decisions travel)
  • LOW security logged in BUGS-FOUND 2026-11-11 (profile-state.jq newline forge, display-only); the fix is tracked there, not here
  • CHANGELOG [2.0.0] (done ed67c52, fixed in place) "Linux make test still to be done after 2.0.0" is stale: suite green on Linux (bchanot-server, 2026-10); note it under Unreleased
  • BLK-018 decided 2026-11-02 (phase 2, option b): /release-candidate runs gitflow.sh finish and the tag in the main loop under the ask prompt; release-executor keeps prep only (536d39e)
  • CHANGELOG internal history headings renamed to [internal X.Y.Z] so the public [3.0.0] is unique (user 2026-11-12, chore/reconcile-pre-3.0.0)

Docs backlog (doc-sync 2026-10-06)

  • P17 README restructure to Standard-Readme (Install / Usage / Configuration / License); move inline reference parts to USAGE.md, CONFIGURE.md from settings.json + .env.example
  • P18 README Requirements line (Linux apt/dnf/pacman + macOS brew): Linux run done, write the line
  • P31 USAGE narrative: Exemple 9 "sans superpowers" (vendored, always on), "Edit tool" bypasses /hotfix, Pattern E + Exemple 9 route bugs to /ship-feature instead of /bugfix|/hotfix
  • P32 USAGE missing sections: design work (21st CLI, design stack, /site-motion), media generation (Higgsfield toggles), gitflow auto-push + global hooks
  • P33 USAGE per-pattern token figures have no source in code (USAGE.md:363,369,499,552,650,712,1019) — verify or drop; "Budget Pro ~11k" already dropped by the D3 add-ons (9b12b12)
  • P34 done by D3 (9b12b12): 0 "gstack ON/OFF" / "context7 ON" left in USAGE.md and agents/plugin-advisor.md; onboard/tour leftovers logged in BUGS-FOUND
  • P41 (2026-11-02): ask prompts under auto mode (bench + interactive check), so the template ask list is live, not inert; README and SETTINGS.md corrected (0bf4620)

User decisions and hand edits

  • user decision: pin a commit for higgsfield-ai/skills and a version for @higgsfield/cli (security gate, 2 MEDIUM, accepted as is)
  • user hand edit pending: remove the four Bash(npm … -g|--global *) lines from permissions.deny (they override the classifier); then the first global install is the live test of the entry

Blocked

  • G2 hooks/guard-bash.sh (BLK-022, safety classifier withheld the body): PreToolUse Bash, exit 2 + reason, logger trace, fail-closed without jq; spec lib/tests/guard-bash.test.sh (214 cases) SKIPs until it exists
  • launch claude with TMPDIR=$HOME/.cache/claude-tmp (BLK-021 resolved by lib/cc.sh, phase 5; .bashrc swap done, ~/.bashrc:133)

Model-router live checks (opportunistic, need a real session or incident)

  • LIVE CHECK after /reload-plugins: best-tier phase w/ Fable out of credits → /route show lists fable down (probe), main on opus, effort bumped.
  • LIVE CHECK: does parentModel follow the turn.step rewrite or the session model? Read general-purpose's step-0 message.model in a transcript during a demoted turn (if it is already the session model, the pin is a harmless no-op).
  • W1-C live verification part 2 (needs a real incident): StopFailure vs turn.complete order; PostModelSwitch auto semantics and its from_model after a router upgrade; whether a router rewrite raises auto; [1m] carry validity on opus (only on a session whose steps carry it)
  • W4-B AC9: type a few plain prompts, read route: classifier <phase> (verbose line with /route verbose on), drop bookkeeping if it misfires
  • W3-A: a parallel same-agent dispatch answered after > 10 s; what an unanswered dialog resolves to on the terminal (must be Later or nothing written). T1 skill-row dialogs seen live (f8665b2, d3635b6)
  • cold-read cap on the opus session: a coding prompt → spinner sonnet-5-5/medium; above 120k cold read the skip line once; /route show cap text

Optional backlog (user call)

  • router: /route set, /route confirm, (unconfirmed) in show, dialog edits of phases, frontmatter auto-alignment, session-start warning when routing.json is dirty, per-machine projects.; step heuristics, quota-aware downgrade, A/B
  • gates.sh could refuse a CHECK holding <<
  • per-skill positive/negative prompt ranking (Tier 2 second half), not started
  • frame floor-guard snippets as data in the verifier step (LOW)
  • graphify pilot on robin_petier: graph + .graphifyignore + gitignore graphify-out/ + GRAPHIFY_FORCE=1 graphify update . in the post-commit hook when a graph exists

Parked

  • Higgsfield parked (final review, rulings in BDR-109): remedy line ignores a pinned lock version (latent while latest); Step 8.6 spawns higgsfield version up to 3 times; rollback needs npm uninstall -g @higgsfield/cli + session removal + hand removal of skills-external/higgsfield-*
  • Higgsfield parked (per-task minors, none blocking): "rename" comment vs rm-then-mv; no -- before the clone URL; ssh URL can prompt; no sweep of a stale .higgsfield-stage.*; timeout path itself untested; "pack not installed" when only unlisted skills are synced; doctor version read unbounded
  • Reticle pilot recipe: section "2026-09-27" below (trigger: first app-type project, BDR-103)

Known limits (accepted, no action)

  • model-router spawn path unprobed: 1st gate of an outage window dies once (seen live 2026-10-23, retry lands on the opus floor); tour/onboard/audit-delta no retry; onAutoSwitch with main on opus marks nothing if the engine auto-falls a gate off Fable
  • Residuals (accepted): explicit Agent(model:"fable") pins (client-handover) still fail w/o credits (BDR-115 kept); spawn path not probed; session model itself out of credits → CLOSED 2026-10-16 c11cd3b (kept cur probed too); no cross-session persistence → ≤8 s stall per new session; timed-out probe keeps retrying in engine (no cancel); [1m] gating unverified (bare id probed); no Esc regression test (kit cannot abort signal); probe sleep eats ~8 of 10 s turn.step budget.
  • Residuals left (accepted): TTL known only after the first PostModelSwitch of the process (no API exposes the account TTL; 60 min default before); hooks-below context and a dropped prompt leave the estimate conservative/stale; onAutoSwitch targets main's lastPlan even for a sub-agent's auto fallback (pre-existing); agent_id: null on a main switch (if the engine ever sends null) would freeze sessionModel/TTL updates.
  • W1-C accepted-by-design (security 2026-10-09, MEDIUM, not coded around;): (1) the model itself can raise the main loop to fable for the rest of a turn through the route tool (plan/reflect/escalate/judge) or a background dispatch (derived orchestrate = best tier) — bounded by the turn and upgradeMaxTokens, no sticky route is model-callable; (2) CLOSED by feat C 786ee13 (keywords, ultrathink and derived orchestrate set effort only). Residuals: PostModelSwitch auto covers "other programmatic change" (a healthy model could be marked 15 min); strikes never decay inside a session; a [1m] variant's model_not_found marks the base model until reload; classifier deferred to W2. 2026-10-23 feat C 786ee13: (1) derived-orchestrate part and (2) keyword/ultrathink part CLOSED (effort-only routes); (1) route-tool lift stays by design (sanctioned path)
  • W3-B residuals: SKILL.md symlink follow (LOW, kept: skills/ itself is a symlink); descs cap and "(custom)" label untested. (security-auditor changed entry purged 2026-10-11.)
  • W4-B: W4-B residuals: smallModelBusy assumes the engine's small model is the haiku alias ($.model.classify takes no model: documented in code 2026-10-11, no fix possible); abort paths judged by reading only; the 3-s tests pin the upper bound only.
  • W3-A residuals (security LOW, accepted): non-atomic cross-process write (two sessions, crash mid-write → file reads as failed, previous config kept); host/path of a private remote in the tracked file; a persistent write failure re-asks every use (asked.delete in the catch); out.length vs byte size at the cap; changePatch scope 'project' with a vanished key writes Everywhere (cwd change mid-dialog).

History (2026-10-08 → 2026-11-12, done)

2026-11-12 — BDR-125 context mod queue shipped (phases 0, 4a, 1, 5, 2 shipped; 3 closed; 4b-4d proof still open above)

  • Before phase 0: /feat /model scope dialog in model-router (done 2026-10-25, 1fde5bb + docs 773fe17, BDR-126; merge on user signal; live check DONE 2026-10-29 (/model opus → scope question rendered → This project wrote settings.local.json); still unchecked: no PostModelSwitch/down mark after This project, gitignore of a created file). (user pick 2026-10-25). Fact (docs model-config): /model <name> always writes model to user settings.json (tracked here); picker key s = session only; project model honored at launch. Mod: command.run on model with args → $.ui.ask "This project / This conversation / Everywhere": project = model in .claude/settings.local.json + sticky route now; conversation = sticky route only (/route model=); everywhere = native passthrough. No args → native picker untouched; -p or no surface → passthrough. Files: register.ts, register.test.ts, lib/effort-shift.md. Doctor keeps the tracked ≠ opus warning.

  • /bugfix model-router haiku effort strip (done 2026-10-25, 8e93507; merge on user signal) (BLK-035 residual, user pick 2026-10-25): Haiku 5.5 takes effort; drop 4.5-era "haiku takes no effort" (effortWord, step plan omission, isHaiku sites, census lib/tests/effort-routing.test.sh:170), so mechanical runs low.

  • /feat picker scope (done 2026-10-25, 2202964 + docs 5d262d3, BDR-127; live check DONE 2026-10-29 (Sonnet 5.5 pick held by the hook, scope asked; same-model pick fires no hook → BLK-037, fixed by BDR-132 657e5cb); still unchecked: PostModelSwitch on a denied pick) (BLK-036, user ask 2026-10-25): bare /model → pick in native picker → ask project scope (.claude/settings.local.json) or user scope. First: engine types for a settings-write intercept; none → mod restores user settings.json = BDR-116 amendment, user decides before code.

  • Phase 0, no mod: split project CLAUDE.md (this repo + the lever, feature/claude-md-split, 2026-10-26; merge on user signal). Four rules under .claude/rules/ (rules-maintenance, graphify-vendored, mods, transient-artifacts # static-check: skip), lib/project-rules.sh check/matches/sections/extract/emit (-C <dir>) + suite, hooks/instructions-loaded.sh (log ~/.claude/logs/instructions-loaded.jsonl, session + agent ids), !.claude/rules/ here + socle, /onboard split + onboarder MODE: split, generators emit docker/testing/ci from templates/project-rules/. Baseline 2026-10-26 /context: 32.2k total, Memory files 11.6k, project CLAUDE.md 2.1k.

    • Proof, this repo (journal 2026-10-26: headless 17:00:07Z 4 Project path_glob_match, sub-agent 17:00:33Z loads rules too; /context after merge = the 2026-10-28 baseline 30.5k; verified 2026-11-02 /reconcile). Was: /context after the merge; one fresh session Reading one file per rule → 4 Project path_glob_match lines; one sub-agent Read → does a path rule load inside a sub-agent. Results in the journal.
    • User step: claude mcp remove -s user higgsfield (done: ~/.claude.json mcpServers empty 2026-11-02; higgsfield is a claude.ai connector, denied per profile since phase 5).
    • Phase 2 folds: narrow the pre-commit .claude/** exemption so a rules-only commit cannot land on develop/main (deferred 2026-10-26; NOT done by phase 2 2026-11-02, still open: its own /feat on the generated pre-commit). Done 2026-11-06 c5bd0f8 (BDR-139): allowlist memory/tasks/audits/deploy + .githooks, --no-renames.
  • Phase 0b, per-repo pass (2026-10-27: 10 repos classified; split on branches awaiting the user's merge: zenquality 1942dcf 4 rules 46%, game 527bdcc 2 rules, angelscar356 ab75ecd 1 rule; keep-everything verdict, no branch: config, integration_alphalink, robin_petier, trucks_at_work, bchanot-cv, for_you, nuit-folle (all sections always-on by the S4 rule); faunosteo 7a75974 11 rules on the user's feature branch, 31.9k → 27.2k; zenquality/game/angelscar356 merged 0ff05f1/d9be6b1/853b55b; proof per repo = the user's next session there) (after phase 0 merges; via /onboard split, one chore/claude-md-split per repo off develop, approval per repo, never an automatic merge; /context before/after in that repo's journal). Order by size: faunosteo 52,322 chars (on feature/v1-1-clients-animals with 5 dirty files: finish or stash first, S1 refuses otherwise), zenquality 23,378, game 14,918 (on feature/replay-hint), robin_petier 12,895, config 4,990, integration_alphalink 4,249. None has .claude/rules/ yet; all use the .claude/* file form (BDR-003). Sections a sub-agent needs stay always-on unless the phase 0 sub-agent proof says path rules load there (proof 2026-10-26: they do; verify per repo).

    • Before faunosteo: ### granularity shipped 2026-10-27 (feature/rules-h3-extract ab5049e: extract '### title', sections --depth 3, split rows ### <title>). Was: decide ### granularity (STEP 6 review 2026-10-26).
    • /feat extract-hardening (2026-10-27, 14a93e3: _rules_out_path_ok refuses absolute, .., existing or dangling-link out; mkdir and rule-write rc checked before the source rewrite; S0.3 prose in the lib case order; 139 checks). Was: /hotfix after 0b, lib/project-rules.sh extract (security gate 2026-10-27, pre-existing, non-blocking): check the > "$out" group and mkdir -p rc before rewriting $src (today a failed rule write still removes the section, rc 0); [ -e "$out" ] || [ -L "$out" ] and refuse ../absolute <out>; onboarder S0.3 says ####x is a bare title but the lib rejects it (align the prose). Its path-bound content (i18n, Schemas, Stores, PDF, Supabase, Storage, 12.6 KB) sits in ### subsections of the always-on ## Project conventions; ##-only extract moves ~1.5% of the file. Option: extract accepts a ### heading (section ends at the next heading of the same or higher level), sections --depth 3. User decides; game also has a ' in ## 4. POINTS D'ATTENTION… (S4 refuses it until renamed, or extract by index).
  • Phase 4a, memory v2 write side: shell, no mod (shrunk 2026-10-25: the measured recall channels never read Meta; the measured status bug is a reconcile parser bug)

    • Done 2026-10-28 on feature/memory-v2-write (a423594..aab9e4d, registries v2 9698473): BDR-129, LRN-236, LRN-237, EVAL-052. memory suite 202/0, make test green.
    • Phase 4a per-repo pass (2026-10-30/31): 12 repos v2 on chore/memory-v2 (config, integration_alphalink, robin_petier, zenquality, game, trucks_at_work, bchanot-cv, game-deploy, gitea-deploy off develop; angelscar356 off main; for_you, nuit-folle off master). game: 16 headings demoted in place; trucks_at_work: 3 bullet Indexes dropped + BLK-001 resolved; robin_petier: BLK-013 resolved + BLK-015 superseded (closures only in the v1 Index). BLK-038 (upgrade cemented prose false-closed) fixed 2dae528 before game/zenquality; user statuses: game BLK-021 open, BLK-026 resolved, BLK-030 upstream, zenquality BLK-006 open. Replay oracle: the 10 earlier repos byte-identical under the fixed lib.
    • Phase 4a parked minors, code half (/feat memory-v2-minors 2026-11-01, e4fb8f7 + 48ade8f: negated status words, dead empty steal, body ## refused at add, _mem_blob rc 1/2, rc 6 names dropped partial stages, reflow; A2 Index comment-edge + A7 rc 6 upgrade wording dropped as moot, 0 ## Index left outside faunosteo). Doc half still open, next line. Original ledger .superpowers/sdd/2026-10-27-memory-v2-write/progress.md: empty memory.lock.steal never stale (manual rm); ## Index comment-edge mismatch; upgrade header doc lacks the non-table refusal; memory-commit rc 6 message misleading on a v1→v2 heading-rewrite upgrade; labelled "not resolved yet" reads resolved on v2; body ## line splits an entry (refuse at add); -C + relative --body-file (say absolute in § WRITE); tour/deploy/audit-delta § WRITE pointer; reconcile.sh:24-25 and :41-42 stale comments; _mem_blob → git cat-file blob; evals template "Later status" line; partial stage dropped on rc 6; 39 pre-existing >80-char lines in reconcile.sh / memory-commit.sh.
    • Phase 4a minors, doc half (/feat memory-v2-minors-docs 2026-11-01): § WRITE intro says every verb takes -C and --body-file is absolute; tour/deploy/audit-delta point at § WRITE; evals template Later status line; rc 6 text done by the code half's doc sync (18e6370).
    • memory.sh awk operand (security LOW 2026-11-01): a --body-file named x=1.md is read by awk as a variable assignment in _mem_fence_open and _mem_body_heading (stdin read, check bypassed; cat still inserts the body). Fix: awk '…' < "$f". Also _partial_stages passes the path to git ls-files -s -- as a glob pathspec (quote with :(literal)). Done 2026-11-06 9b3961e (_mem_path_safe at entry points, LRN-265).
    • Facts 2026-10-24 (scratchpad memeval, 434 entries): registries ~1M chars, never in always-on context, read only by /feat /bugfix /ship-feature (analyze-before-plan); Index drift 6 rows; status in 3 places (title tag, body Status line, Index column) → reconcile false-open BLK-032..034; no single writer (Edit needs a Read of a 357k file, so real writes are ad hoc python/bash); 8 projects on this machine hold v1 registries. Verified 2026-10-25: lib/reconcile.sh:60-67 drops the entry's OWN header while BLK-032..034 carry [resolved] there (the false-open); nothing machine-reads ## Index (analyze-before-plan.md:42, reconcile.sh:9, test T1); 1 of 1,290 entries across the 8 projects above a 1% accent ratio.
    • lib/memory.sh = single writer, like gitflow.sh: add <kind> --title --body-file [--status --supersedes] allocates next ID from headings (never the Index) under flock (concurrent conversations, phase 5), appends the entry; status <ID> <new> appends a dated status line (append-only); journal "<line>"; index prints the Index on demand from headings; check [--staged] = the validator (heading pattern, ID = max+1 and unique, no removed line from an existing entry outside a /prune-memory commit, new BLK quotes ≥1 error verbatim). The validator runs in add and check + make test, NOT in the generated pre-commit (undecidable there, and it exits 0 on working branches). No English check (doctrine stays).
    • Status: reconcile_blk_current_status reads the own-header [tag] first, then "last status-bearing line wins" (one parser, shared by memory.sh status and reconcile). Fixes BLK-032..034 false-open; regression test on those three.
    • Index: the four generated ## Index blocks are REMOVED in one /prune-memory curation; Index steps dropped in capitalize, ship-feature, commit-changer, prune-memory § D; memory.sh index replaces them.
    • Per-file v2 marker (first line) so the 8 v1 projects keep working untouched; doctor warns "registry v1".
    • Dropped: Meta line, 434-entry backfill, scope/critical labels, triggers field (4c derives triggers at read time), Index == generated validator, English validator.
    • Citers: capitalize, close, prune-memory, CAPITALIZE steps of feat/bugfix/hotfix/ship-feature/commit-change, lib/capitalize-commit.md, lib/analyze-before-plan.md, lib/reconcile.sh, templates/memory, onboard/init-project; doctrine-citers census. CLAUDE.global.md Memory registries: write only via lib/memory.sh; fix the false "load every session" premise (keep caveman: it keeps reads and injections small).
    • Proof: writer round-trip, flock race test (two adds), status parser on the three false-open BLKs + fixtures, v1 project untouched, make test green.
  • Phase 1, native context diet: no mod (redesigned 2026-10-24, shrunk 2026-10-25: one carrier per section, existing files first)

    • Done 2026-10-29 on feature/context-diet (4bbb3c5..789a2b2): BDR-130, BDR-131, LRN-238..241, EVAL-053. /context 30.5k → 26.2k (own −2.8k). Verifier CONFORME 21/21, security PASS. Contract .claude/tasks/contracts/2026-10-28-context-diet-1944.md § PROOF.
    • User step: delete the auto-memory leftovers by hand (done: the three dirs are gone, every ~/.claude/projects/*/memory/ empty 2026-11-02).
    • Remove the dead memory: project field from agents/analyzer.md (/hotfix 7899670, 2026-10-31).
    • Census blind spots (2026-10-31, feature/citers-blind-spots 5bd0de4 → develop 4175401, T8a-T8h): nested lib/<dir>/x.md § Y, templates/ and .claude/rules/ outside doctrine-citers. Measured 2026-10-31: 0 real citations in all three today (preventive).
    • Measure the design hook's "Read … first" re-Read cost (2026-10-31, transcripts since 789a2b2): 59 hook fires, 0 Read-tool reads of web-building.md, 8 Bash mentions (repo maintenance). Re-read cost ≈ 0 → no "once per session" needed.
    • Probe Bash sed -n (2026-10-31, two claude -p --model haiku runs here, mods/model-router/.claude-plugin/plugin.json): Bash sed -n AND Read both log path_glob_match mods.md with that trigger file. Path rules load after a Bash read too; not isolated whether the harness or the rtk rewrite triggers it.
    • Minor: relative rule paths in interviewer.md / guardrails SKILL.md → ~/.claude/rules/ (/hotfix 40a993c).
    • Agent description losses (3330f0b): user pick 2026-10-31 = restore "(dispatcher applies)" on geo-analyzer + seo-analyzer only; the rest closed (bodies keep it).
    • robin_petier chore/memory-lan-server (BDR-075): merged 2026-10-29 (6877443).
    • Facts 2026-10-24 (scratchpad pathtest, memtest): a paths: rule loads on Read of a matching file, NOT on Write of a new file; "autoMemoryEnabled": false removes the auto-memory system section; classic SessionStart / UserPromptSubmit hooks inject additionalContext; instruction files reload on compaction natively. hooks/design-toolchain-reminder.sh fired 2,131 times (16 to 210 per day); brainstorming is vendored and gitignored (.gitignore:78), an edit there does not persist.
    • Section → carrier:
      • auto-memory system section → "autoMemoryEnabled": false in settings.json. Before: migrate existing auto memories in faunosteo, jeux-deploy, robin_petier into their registries (user review), doctor warns when an auto-memory dir is non-empty.
      • Design work (1.5k chars) → APPENDED to rules/web-building.md (same UI globs, one file); hooks/design-toolchain-reminder.sh keeps its one-liner, retargeted to "Read ~/.claude/rules/web-building.md § Design work" (covers new-file turns); never the full text per fire.
      • graphify → one always-on threshold line stays ("graphify from 200 code files, never below"); the rest of the section → hooks/session-start.sh injects it only when graphify-out/graph.json exists.
      • Public websites never SPA + default language + web APIs versioned (2.3k) → one lib/architecture-defaults.md Read by our own skills (init-project, ship-feature STEP 1, refactor whole-project, onboard; the lib/*.md reference pattern); Web APIs § merged into rules/rest-api.md; no empty-dir injection (init-project routes correctly in an empty dir, routing probe C1).
      • secure-by-default: "Every API", hosting, injection, exposure, SMS/email stay always-on; Supabase and cloud functions keep ONE always-on line each (path rules fail open at scaffold/Write), their bodies move to path rules (supabase/**, **/*.sql, **/migrations/**; functions/**, **/serverless.*, **/firebase.json).
      • Agent descriptions: trim description: in agents/*.md at the source (5,755 of 9,098 listing chars; names and files unchanged, dispatch cannot break). Replaces the phase 3 agent listing rewrite.
      • Registry format → compressed in phase 4a; gitflow detail → compressed in phase 2 cleanup. Our-code-vs-foreign line: already injected by session-start.sh:240-251, nothing to do.
    • Proof: static glob check per moved rule + logged session; hook tests (design-toolchain-reminder one-liner, session-start graph branch); the four skills Read the lib (grep census); /context delta.
    • Cleanup: CLAUDE.global.md loses the moved sections (one pointer line each where needed), citers patched (doctrine-citers census); BDR-031 citers.
  • model-router: ask before routing to Fable (done 2026-11-02: feature/router-ask-tier 9a24a62 + docs 789fd00, merged d4416fe; BDR-135, LRN-254; GATE 0 MET, verifier CONFORME 11/11, security PASS; contract .claude/tasks/contracts/2026-11-02-router-ask-tier-1434.md; live check pending: a typed best-tier skill from an opus session shows the question) (user ask 2026-11-02, after the phase 5 merge: both touch mods/model-router/routing.json). /feat on the mod: a routing.json key askTier (default best; /route ask-tier on|off) makes the router ask, once per span, BEFORE applying a route whose resolved model sits in that tier, with the reason (phase + its about, origin: skill row / /route / prompt rule / agent row) and the options Fable · Opus instead (big tier) · stop asking this session; no memory in confirmed (the existing first-use confirmation stays as is); never asks in -p (askOr already falls back to LATER). Files: mods/model-router/hooks/register.ts (+ register.test.ts), routing.json, USAGE.md (/route section).

  • Phase 5, per-project and per-conversation profiles, native packs (redesigned 2026-10-25, shrunk the same day; shipped 2026-11-02 on feature/phase5-profiles-packs, merged 3cc4e0d: BDR-134, contract .claude/tasks/contracts/2026-11-01-phase5-profiles-packs-1300.md; spec and plan purged at finish (BDR-065), facts in LRN-248 and the contract; gates re-run 2026-11-02: 7/7 MET)

    • User steps (verified 2026-11-12: ~/.bashrc:133 sources lib/cc.sh, no dtach_claude/alias cc; /context full/dev/web recorded in phase 3 below): MIGRATION "Profiles and packs (2026-11)" step 4 .bashrc, then /context under cc t full, cc t dev, cc t web.
    • Facts 2026-10-24/25 (scratch projects, fresh headless sessions, canaries, run concurrently, one shared ~/.claude): .claude/settings.local.json enabledPlugins per project works both ways and concurrently; claude --settings '<json>' gives two SIMULTANEOUS conversations in the SAME project different plugin sets; a skills-dir plugin obeys both; plugin skills keep their short name (ui-styling resolves) but verify <plugin>:<name> too (router rows must match both); plugin agents need the plugin: prefix; permissions.deny Skill(x) does not hide a skill; skillOverrides does not reach plugin skills (packs hide by disabling, guide); a user-scope MCP cannot be switched off per project by settings: project MCP belongs in a pack .mcp.json; claude.ai connectors do not load under -p.
    • User answers 2026-10-25: profile per project and per conversation, never one for all; tracked .claude/settings.json keeps deny/allow only, profiles go local; a project without a profile gets a proposal once, written on accept; the global mode (skills/ ↔ skills-disabled/ symlinks + user-scope plugin toggling) is DELETED; core = everything except the heavy packs (design, seo/geo, web-validate, scroll/3d, higgsfield, gstack); agents stay global; a skill needing an off pack proposes to enable it then /reload-plugins (spike first); all 11 profiles kept as pack compositions.
    • Packs: each heavy group becomes a skills-dir plugin skills/pack-<name>/ (.claude-plugin/plugin.json + skills/<skill>/SKILL.md, + .mcp.json for higgsfield, + hooks if any). Spike first: nested skills load once as plugin skills; gstack (external installer, symlinks) as a pack.
    • Model per project and per conversation (user ask 2026-10-25, measured headless the same day: project .claude/settings.local.json "model": "haiku" → haiku; --settings '{"model":"sonnet"}' on top → sonnet; nothing → the user settings model; scratchpad modeltest). A profile carries model beside enabledPlugins; profile.sh set writes both into .claude/settings.local.json; the conversation launcher passes both. /model writes the TRACKED user settings.json (seen 2026-10-25: fable landed there), so the doctrine says: /model only for a one-off turn, the project model goes through the profile; the tracked default stays opus; doctor warns when settings.json model differs from opus. Applied by hand today for this repo (local fable, tracked opus).
    • Profile = lib/profiles/<name>.json ({"enabledPlugins": {...}}), the source of truth, reachable through the ~/.claude/lib link: no generator, no ~/.claude/profiles/. Per project: profile.sh set <name> writes enabledPlugins into .claude/settings.local.json (never user scope). Per conversation: claude --settings ~/.claude/lib/profiles/<name>.json (shell function or make helper, named with the user).
    • Detection: no fingerprint. When the project has no profile, session-start prints one line pointing at /plugin-check; the plugin-advisor maps its canonical profile to a pack profile (replaces the banner hint the cleanup deletes).
    • Pack needed mid-skill: the orchestrators' existing gates (lib/design-gate.md + design-tool-gate.sh, lib/plugin-gate.md) check the required packs; off → propose enabling for THIS conversation (the --settings copy when the session was launched with one, else .claude/settings.local.json; never user scope) + /reload-plugins; spike that a settings change + /reload-plugins loads the pack mid-session.
    • Listing budget (user pick: after the packs): 124 skills listed, 46 name-only at the default budget; SLASH_COMMAND_TOOL_CHAR_BUDGET=60000 → 0 name-only (+~5.7k tokens). Once packs shrink the listing, set the budget in settings.json env to the smallest value that leaves 0 name-only in the full profile (ctxprobe bench), record the cost.
    • Spikes: nested skills in a pack, /reload-plugins after a settings change, claude.ai connectors per project (interactive only), higgsfield MCP in a pack.
    • Proof: two concurrent conversations in one project and two projects, canaries prove different listings, MCP and hooks per pack; /context per profile; nothing references skills-disabled/ or toggle-external.sh.
    • Cleanup: lib/profile.sh symlink toggling + skills-disabled/ (24 entries), user-scope plugin toggling in set|apply|reset, lib/toggle-external.sh (retire or rewire to packs) + lib/tests/higgsfield.test.sh + the CLAUDE.global.md "toggle-external.sh enable higgsfield" lines, every writer of the old layout (install-plugins.sh, update-all.sh, gstack and 21st installers), Makefile profile* targets rewired, /profile skill rewritten for packs, plugin-gate prose tied to global toggling, banner hint "/plugin-check before starting a new project", doctor profile checks, profile tests rewritten; lib/profiles/*.profile → .json.
  • Phase 2, guards (shipped 2026-11-02 on feature/phase2-guards, merge on user signal: bench .claude/tasks/plans/2026-11-02-phase2-bench/ (16 ask, 7 deny approved; quoted finish rule, gitflow_finish*, bash -s* not approved), rtk rc 3 passthrough 60ee4a9, router write guard with a session skill latch 22ff57b + 9bbd1d5, code-limits hook f21e265, locks T62/T63 9887c14, settings 6899a72, release finish in the main loop 536d39e, doctrine 0bf4620, CHANGELOG c3a6977; contract .claude/tasks/contracts/2026-11-02-phase2-guards-1545.md) (redesigned 2026-10-25 after a 3-lens challenge, re-challenged the same day: native ask gate, guard hosted in model-router, interpreters in prose, commit-msg hook deferred)

    • User hand edit (done 12c3534, merged 29cb080; CHANGELOG :51 14 entries / "Edit of" + Fixed bullet :92 + "(98 checks)" dropped in ed67c52, merged 1b98170): the 6 dead deny entries removed from settings.json; push-guard T41/T42 lock prepared by feature/dead-deny-lock 79f1a3f.
    • Killed by the challenges: the phrase-list merge gate; the $.ui.ask mod gate (its BDR-068 exception detector was never true: finish runs in the user's reply turn; a hard_deny naming mod sources was the robustness BLOCKER); G2 destructive guard (BLK-022, stays a user-driven task); whole-file code-limit reports (1498 lines > 80 cols across 92 files, over-fires); parsing commit text from Bash; Bash write-form detection in the routing guard (no measured Bash evasion, every inline write was Edit/Write).
    • Merge gate = native ask rules in settings.json, no mod (user pick 2026-10-25; bench r3 G1-G7: all forms DENIED headless incl. "$HOME", absolute path, after source; start RAN): Bash(bash */gitflow.sh finish*), Bash(gitflow_finish*), Bash(git merge *); evaluate in the same bench Bash(git pull *) and git push x:develop forms before adding them. Trade-off accepted: the BDR-068 /close auto-finish gets one dialog click (ask beats allow, no exemption); release-executor finish prompts too (its dispatcher runs finish inline if a sub-agent cannot prompt, BLK-018 option b). hard_deny line: "never edit permissions.ask to skip the gate".
    • Routing guard (user pick: "always through a skill") hosted in mods/model-router (owner of the only run signal; mod state is plugin-private, runMain is BEST-tier-only and never ends: wrong signal). tool.call on Edit/Write/NotebookEdit of a code file by the MAIN loop (no agentId) → deny with reason "route through /hotfix (≤2 files, obvious), /feat, /bugfix or /ship-feature" UNLESS this turn's turnMain.source ∈ {skill, model} or a /route sticky is in force (explicit escape hatch). Free: docs (.md), .claude/**, sub-agents, any write inside a skill run. Bash write forms deferred until a measured evasion. hard_deny gains "a guard redirect is not a refusal to route around: enter the skill". Doctrine: the ceiling bullet becomes "the main loop writes no code outside a skill run, whatever the model". Proof: routing bench cases (L3 must enter a skill), long-brief replay in a clone, false-deny fixtures (doc edits, executor edits, in-skill edits), guard-off = today.
    • Shell deny hygiene (bench .claude/tasks/plans/2026-10-25-bash-deny-probe/; user pick: stay in defaultMode: auto). Facts: the native matcher sees through && ; || | newline ( ) env command \cmd timeout nohup xargs quoted command word and simple assignments; it misses sh -c, bare bash/sh + heredoc, $(…) in command position, absolute paths, interpreters and a quoted literal argument; under auto every miss was approved by the classifier; the rtk-rewrite hook runs BEFORE permission evaluation and deny is evaluated on the REWRITTEN command (git push origin :feature → rc=3 → rtk git push … bypasses deny git push * :** and ask git push *; rc=0 misses cat */*.pem); redirects into an Edit-denied path ARE denied; ask beats allow. Counted 2026-10-25 (14 days): 211 python3 -c, 450 node -e, 491 of them from sub-agents: a static ask on interpreters would prompt or deny hundreds of calls.
      • rtk hook: rc=3 → pass through unchanged (no updatedInput), native deny/ask see the original; re-pin hooks/.rtk-hook.sha256; hook test with the cat SECRET* analogue hooks-on. rc=0 reads: census test that every Bash(cat|head|tail <glob>) deny has its Bash(rtk read <glob>) twin, twins added (alternative weighed and rejected for now: one case arm passing cat/head/tail through unrewritten loses rtk's read savings, unmeasured).
      • settings.json ask += Bash(sh -c *), Bash(bash), Bash(sh), Bash(bash -s*), Bash(bash <(*), Bash(/bin/sh *), Bash(/bin/bash *), Bash(/usr/bin/env bash*), Bash(/usr/bin/env sh*) (NOT Bash(/*): it would kill the Bash(*/rtk …) absolute-path twins the rtk heal relies on); deny += Bash(awk *system(*. Interpreters (python3 -c, python -c, node -e) go to the autoMode soft_deny prose, not static rules. Each pattern proven in the bench before landing. P41 (templates: "ask inert under auto") is contradicted by measurement: fix the item.
      • Limits written down, one place: settings.json soft_deny note + README permissions: $(…) in command position, a quoted literal argument and interpreters have no static rule and rest on the classifier + prose; mode stays auto.
      • Folds TODO Actionable "docker host-root mount spellings" and the T42 dead-entry drop list if the release lands first.
      • Interim (accepted): until this lands, a git push origin :ref deletion rests on the classifier and rtk's registry.
    • Commit message hook: DEFERRED (4 of 78 non-merge commits over 50 words since lib/commit-message.md landed e91dcb3; the generated hooks miss 6 of 7 repos on core.hooksPath=.githooks). Revisit if the overrun grows.
    • Code limits: classic PostToolUse hook, added lines only (Edit new_string, Write content), our code only (lib/code-origin.sh), 80 columns only (462 of 19,161 added lines measured over), at most 5 reports, once per file per turn, additionalContext, never a deny.
    • Proof: evasion matrix per rule (bench), the /close ritual finishes after one click, destructive commands never run live (plugin test with stubs only).
    • Cleanup: compress the gitflow section only for the forms now enforced.
  • Phase 3, context mod: CONDITIONAL — CLOSED 2026-11-03, not built (measured by the user, repo root, Opus 5.5, after phases 0/4a/1/5/2: full 40k = memory files 8.2k, skills 16.5k, system tools 7.6k, MCP instr 2.4k, agents 2.2k, system 2.2k; dev 38.4k = memory 8.2k, skills 12.7k, system tools 11.5k, MCP instr 0.7k; web 40k. Connector instructions already have a native lever (profile deniedMcpServers: 2.4k → 0.7k on dev); the jit-subtractable doctrine left in memory files is ~1-2k (3-5%), not worth a mod birth, canaries, off switch and doctor line.) (user pick 2026-10-25). Measure the always-on residual after phases 0, 4a, 1, 5 and 2 (/context at repo root + the two biggest projects). Build only if a mod can carry conditionally a residual worth its cost (birth, off switch, canaries, doctor line); otherwise close the item with the measurement.

    • If built, scope = prompt.compose section deltas for doctrine a carrier already re-delivers (jit-marked, never deleted; mod off = full text; ledger in $.state, reset on compact/clear) + mcp_instructions_delta for connectors irrelevant in a coding repo. Shared infra (jit-marker subtraction, $.state ledger) lives here or in 4c, whichever ships first.
    • Dropped by the re-challenge: skill listing filter + routing.json contexts + census (phase 5 packs + the budget env already give 0 name-only); routing-hint classifier (no measured miss outside code intents the phase 2 guard catches; the "Skill routing" table lines move into our own skills' description:, ≤3 lines for vendored aliases stay in CLAUDE.global.md, done in phase 1 cleanup); agent listing rewrite (phase 1 diet); higgsfield deltas (phase 0/5); project fingerprint. prompt.attachment kinds (skill_listing, agent_listing_delta…) exist in the 2.1.306 types but not in the docs text: re-verify at build time.
  • Measure the 30.2k → 40k rise (2026-11-09, LRN-270): Skills +6.2k = budget 50000 paying mostly claude.ai-synced skills; System tools rise not attributable (/context all has no per-tool row). Tool-search threshold hypothesis disproved (ENABLE_TOOL_SEARCH=true changes nothing; Higgsfield deny on full keeps 7.7k). L1 shipped: feature/claude-ai-skills-off 47438f4 + 07406ce (BDR-141), full headless 32.2k → 28.5k; user removed Higgsfield + unused claude.ai connectors (full 26.9k headless). Merge on user signal.

    • Cleanup run (2026-11-09, /ship-feature on chore/connector-deny-cleanup b80da29..5413f40, 12 files): denies + census connector rule removed, prose citers rewritten, T13 deny-set variant added; gates MET / CONFORME (3rd pass) / PASS; BDR-134 amended, LRN-272/273. Merge on user signal. Not checked: work machine project settings.local.json (CHANGELOG says delete by hand).

2026-10-19 — guardrails split: /a11y + injection/exposure (feature/a11y-split-inj-exp)

  • run — SDD 9/9, gates 0/1/2 green, opus review Importants fixed (c0fdbec, 9275e0d, 63a88bc), capitalize BDR-122 LRN-219..221 BLK-032; merge = user go
  • parked review Minors + GATE 2 residuals (probe supabase.url, redirects, _nonempty_rows, printable) → /bugfix 2026-10-21, 5a98c2c + 633faa2, BLK-033
  • routing policy /feat (user 2026-10-19, BLK-032): settings model → Opus 5.5; judge split (security-auditor + verifier → Fable, plan-challenger/analyzer/code review → opus); executors stay sonnet, "what to write" on opus; ceiling rule: main on Fable never writes code beyond a hotfix edit → dispatched executor; BDR superseding part of BDR-066 / BDR-115.
    • feat A audit phase: d3635b6 (dialog writes, session model opus), a7d747c (audit row, opus floor, census, citers), dae142b (docs); gates 0/1/2 green, LRN-224
    • feat B (feature/routing-ceiling-rule): ceiling rule CLAUDE.global.md, README doctrine, challenge citers, deploy effort: xhigh, routing.json changed cleared; ef13713 + 887cf89, gates 0/1/2 green, BDR-123; merge = user go
    • feat C: cap the default keyword prompt rules (plan|design|architecture, why|review|audit|analyze, ultrathink) at the session model, effort only like the classifier; close BLK-032 after — DONE 2026-10-23 feature/router-keyword-cap: 786ee13 effort-only routes (keywords, ultrathink, derived orchestrate), d2d590f docs, gates green, BLK-032 resolved; merge = user go
    • bugfix lib/tests/mods.test.sh (bugfix/mods-validate-advisory): verdict = --strict rc, positive control warnfix/cleanfix; 801992a, gates green, BLK-034 LRN-225; merge = user go. Residual: 3 historical contracts keep the old scrape oracle (BLK-034)

2026-10-18 — secure-by-default guardrails: 5 domains + /guardrails skill (feature/secure-by-default)

User ask: hosting, a11y, Supabase, SMS/email consent, cloud functions must be baseline on every project, API and access secured systematically, plus a skill that tests all of it. Language call: scanner is Python stdlib, not Rust (default-language contraindication: skill scripts here are interpreted, no build step, same as explain-diff checker).

  • T1 rules/secure-by-default.md always-on (no paths:, BDR-085 path: CLAUDE.global.md is at 320/320)
  • T2 skills/guardrails/scripts/ scanner: common, apis, hosting, a11y, supabase, messaging, functions + scan.py CLI (FAIL / WARN / MANUAL, exit 1 on FAIL)
  • T3 skills/guardrails/SKILL.md + references/checks.md (rule catalogue, manual items, live Supabase queries) + test-prompts.json
  • T4 lib/tests/guardrails-scan.test.sh: clean fixture passes, one mutant per FAIL rule
  • T5 wiring: security-auditor checklist (+ scanner run), interviewer failure-mode row, init-project founding table
  • T6 registration: profiles full/max/dev, routing.json row, README, USAGE x2, CHANGELOG
  • T7 make test all green (70 guardrails checks), shellcheck clean, 25/80/5/5 limits met; calibrated on 8 real projects (game, faunosteo, zenquality, robin_petier, alphalink, game-deploy, config, explanations), 9 false-positive classes fixed and locked by benign tests. Journal done. BDR-120 accepted, merged bee377a.
  • T8 passwords (user ask 2026-10-18): doctrine bullet in the rule (argon2id, no clear compare, no log, no literal, delegated auth), checks/passwords.py in the api domain (API-PASSWORD-PLAINTEXT / WEAK-HASH / LOGGED FAIL, LITERAL WARN), 8 mutants + 10 benign locks (89 checks), auditor maps PLAINTEXT to CRITICAL; calibrated on 7 projects: 0 hits after one FP fix (yaml volume mount htpasswd:/etc/...)
  • T9 live probe (user ask 2026-10-18, "attaque de l'exterieur ?"): skills/guardrails/scripts/probe.py read-only (GET/HEAD/OPTIONS, --authorized gate, env secrets, same-host redirect only, request budget, 25-req burst cap). 5 probes: anon-read, cross-user, no-auth route, cors-reflect, rate-limit. Suite lib/tests/guardrails-probe.test.sh 16/16 vs a localhost http.server fixture. SKILL STEP 5 + checks.md section + CHANGELOG/README/USAGE. Redirect-guard unit test blocked by the auto-mode classifier (SSRF-shaped fixture), guard is code-reviewed.
  • Fix loop (user ask 2026-10-18): audit -> fix proposals (Fable via route plan) -> user accepts -> fixes applied -> re-pass proves all fixed. Reuse /bugfix circuit: contract with scanner oracles, bugfixer executor, lib/verify-secure-loop.md gates.
  • T10 scan.py --assert-absent RULE:path,... (contract oracle: exit + marker reflect only the listed pairs, positive control = domain applies + files>0) and --baseline <json> (re-pass: fixed / remaining / new); tests incl. positive control
  • T11 SKILL.md: model gate, STEP propose (grouped per rule+path, visible choices named, MANUAL = user steps), STEP approve (prose gate), STEP fix (gitflow aiguillage, contract + oracles, bugfixer, verify-secure-loop), STEP re-pass (baseline diff + probe re-run on go, max 2 rounds), commit gate
  • T12 citers: verify-secure-loop header, contract-interview flow list, bugfixer description, model-routing census adds guardrails
  • T13 docs: checks.md oracle section, CHANGELOG, README, USAGE, test-prompts
  • T14 make test green (scan suite 106: check modes + gates.sh UNMET->MET), oracle calibrated read-only on game: real hit -> PRESENT rc1, self-baseline -> NO-NEW rc0

2026-10-17 — /explain-diff skill: HTML explanation of a diff (feature/explain-diff)

Source: geoffreylitt gist explain-diff-html (Notion variant dropped). Built through the skill-creator loop (draft, evals with/without skill, user review).

  • T1 skills/explain-diff/SKILL.md (target resolution, understand, 4 sections, style, save outside repo, check, hand-over) + test-prompts.json
  • T2 assets/template.html (self-contained skeleton: light/dark tokens, TOC, diagram primitives, diff + quiz JS) + references/components.md
  • T3 scripts/check_explanation.py (deterministic floor: name, location, sections, TOC, 5-question quiz, code escaping, no ASCII art, no em-dash, self-contained) + lib/tests/explain-diff-check.test.sh (17 checks: valid page, raw template, one mutant per check)
  • T4 registration: full/max/dev profiles, routing.json row reflect, README + USAGE x2 + CHANGELOG
  • T5 skill-creator iteration 1 (runs on opus: Fable out of credits): 3 evals x (with skill, baseline), graded, pass rate 100% vs 45%; fixes faedb98 (arrow/edge class collision, bundled fill_template.py). Static viewer in session scratchpad. + iteration 2
  • T6 make test all green, shellcheck clean, impeccable detect [] on the template, contrast >= 5.8:1 both themes; commits b353fac faedb98.

2026-10-16 — impeccable = Build path for all UI (chore/impeccable-build-path)

User validated: frontend-design overlaps impeccable 4.6; gate still blocked on frontend-design. Supersedes the 2026-07-05 "complementary, keep both" call.

  • CLAUDE.global.md Build line: impeccable leads, frontend-design = light path only, rule count 45 -> 61
  • design.profile GATE-BLOCK: frontend-design out, impeccable in (frontend-design stays installed, profile-listed)
  • lib/design-gate.md: core-set prose + section 5 (impeccable now blocking, so only the PRODUCT.md check remains)
  • agents/plugin-advisor.md: frontend/design-system rows, package rows, two RULE blocks
  • 45 -> 61 detector rules: README, plugins.lock.json note, install-plugins.sh comment + echo
  • BDR-119 + CHANGELOG Unreleased + journal
  • make test: 1 red suite = known design-tool-gate precondition (21st CLI installed, see 2026-10-06 entry), shellcheck clean, impeccable resolves under skills/

2026-10-15 — short commit messages (feature/short-commit-messages)

  • lib/commit-message.md: lookup of a documented convention (README, CONTRIBUTING, CLAUDE.md, .claude/ rules, commitlint, template), else 50 words max, medium technical level. Git log is not a convention.
  • Citers patched: commit-changer, bugfix STEP 6, feat STEP 5, hotfix STEP 4, client-handover-writer, CLAUDE.global.md (1 line, 320-line budget).

2026-10-15 — model-router: probe a move target before routing main to it (bugfix/router-probe-before-route)

Contract .claude/tasks/contracts/2026-10-14-router-probe-before-route-0005.md, plan r3 .claude/tasks/plans/2026-10-14-router-probe-before-route-0005.md.

  • /bugfix run: Fable out of credits routed (BLK-030, LRN-218); r1 → r3; 497 tests, verifier CONFORME 11/11, security PASS; commit 2cccc43 + docs.

2026-10-13 — model-router: cold-read cap, switch on by default, classifier routes down (feature/router-downgrade-cap)

Contract .claude/tasks/contracts/2026-10-13-router-downgrade-cap-0013.md, plan r3 .claude/tasks/plans/2026-10-13-router-downgrade-cap-0013.md.

  • /feat run: 3 lenses (10 MAJOR) → r2; confirmation pass FATAL(7) (BLOCKER: served record only on routed steps) → r3; feater DONE (19 tests + 3 extra, 450 pass), GATE 0 MET 4/4, verifier CONFORME 10/10 first pass, security PASS. Commit 2ce203a.
  • Residuals (1)-(3) closed 2026-10-13, commit 1be57a3 (contract .claude/tasks/contracts/2026-10-13-router-cap-residuals-1902.md, plan r3): session pin for built-in general-purpose spawned by main below the session model; prompt-size estimate (text + context + typed skill body, code points / 3) added to the cap's size at the first main step unless usage grew by half; document attachment = unknown = blocked; engine cache_ttl ('5m'|'1h') adopted from main's PostModelSwitch. 3 lenses (2 BLOCKER same finding: frontmatter model invisible to the hook) → r2; confirmation CONCERNS(3) → r3; feater DONE 472 tests; verifier CONFORME 7/7; security PASS. Residual (4) "model unknown to the table" not reproduced since (later route answers named claude-fable-5-1): closed as transient.

2026-10-12 — code-clean whole repo (chore/code-clean)

  • audit r1→r3 (3 lenses + confirmation), user approved all 14 non-deferred items, code-cleaner DONE, make test 47 green — plan .claude/tasks/plans/2026-10-12-code-clean-0103.md, scope .claude/audits/CODE-CLEAN-SCOPE.md
  • fix BUGS-FOUND: toggle-external.sh usage() prints header 3-26, list ends at 32
  • C4/C5 deferred dedupes via /refactor (tool_kind/tool_source + shared gstack_* in lib/gstack-removed.sh; 496-run golden identical) (tool list ×4, profile/toggle-external twins)
  • deferred: C1 gitflow-test.sh move (blocked: T19 drift lock), B13-B17 (no change by design)

2026-10-08 — model-router mod: one mod routes model + effort per request (feature/model-router-mod)

Plan .claude/tasks/plans/2026-10-08-model-router-mod.md. Decisions 2026-10-08: main-loop model switch spike-first then flag off; load via CLAUDE_CODE_PLUGIN_DIRS + link.sh; migration of shifters/pins/model-gate in wave 2 after proof; names model-router / route / /route.

  • W0 spike in dev-mods (hot reload): facts a-d established 2026-10-08 (plan file § Spike facts); e moved to W1.10
  • W1-A the mod in mods/model-router/ (b721c94, contract 2026-10-08-model-router-w1a-1533, plan r3): challenge 3 lenses + 1 confirmation (2 BLOCKER + 10 MAJOR closed by named changes), feater DONE first pass, GATE 0 MET 5/5, verifier CONFORME 6/6, security PASS (4 MEDIUM + 5 LOW reported, below)
  • W1-A hardening (user go "oui durcis", contract criteria 7-11): /route composer-only; no agent model axis (effort only after spawn); pattern ≤ 200 / scan ≤ 4096 / phase keys ^[a-z][a-z0-9_-]{0,31}$ / config ≤ 64 KB / additionalProperties: false / typeof e.skill; warnOnce in all 14 catches + config-drop logs; safely around post-next bookkeeping. feater DONE, gap round (dead Loop.frozen, spawn returns started verbatim, tool description), GATE 0 MET 9/9, verifier CONFORME 11/11, security PASS.
  • W1-A residuals CLOSED 2026-10-11 (bugfix/model-router-residuals): nested-quantifier patterns rejected (NESTED_QUANT); stat.kind !== 'file' → layer failed; rejected keys masked in logs (keyWord), alias keys KEY_NAME; route-tool catch text says the route may be partly applied. Left (accepted): agentId read from the flat tool event; closures without a kit test.
  • W1-B1 floor precedence (contract 2026-10-08-model-router-floor-1835, 2026-10-09): ultrathink + typed /effort-<l> = the main turn's default AND minimum (r2 after 3 lenses: a pure floor made /effort-low a no-op); one helper mainEffort; per-axis precedence; mid-turn prompt floors now + next turn; "enabled": false per machine, kept across /clear and failed reload; typed slash attested. 30 tests; verifier CONFORME 6/6 (after 1 gap round); security PASS ×2
  • W1-B1 residuals CLOSED 2026-10-11: broken override at first load → router off (fail closed, logged); non-boolean enabled → off; /route reload answers config unreadable; previous config kept; errWord (name + code only) replaces every String(err) log. Superseded by W2-B: skill.prompt preload guard, /effort-* marker (bridge removed). Left (accepted): transient missing override lifts a config-set off (ENOENT must mean defaults on most machines); /route show display items (no defect named).
  • W1-B2 wiring (contract 2026-10-08-model-router-wiring-1835, 2026-10-09): tracked symlink skills/model-router → ../mods/model-router loads as model-router@skills-dir (fresh-process claude plugin list --json proves it), .gitignore mods/*/tsconfig.json, lib/tests/mods.test.sh (4 checks, capability probe, bounded, SKIP), doctor ── Mods ── fail-soft, CLAUDE.md ## mods/. Plan r2 from 3 lenses (5 MAJOR), feater DONE, gap round (4b label + unbounded probe), verifier CONFORME 8/8, security PASS
  • W1-B2 residuals CLOSED 2026-10-11: mods suite probe greps the claude plugin test usage line; doctor plugin list --json under timeout 60; update-all skips @skills-dir mods. Left (accepted): doctor echo -e helpers (tracked names only); fallback timeout guard orphans grandchildren without coreutils timeout.
  • W1 close-out (2026-10-09): doc-sync opus audit SIGNIFICANT → user go all 10 → patched (README effort routing + Explore row + /route, USAGE, ARCHITECTURE mods/, CHANGELOG Unreleased) b22f894; hot-reload link removed from ~/.claude/dev-mods/<session>/; BDR-115 amended (a6e2003). Pending user: /reload-plugins in this session (new sessions load the skills-dir copy by themselves); merge decision on feature/model-router-mod (gitflow finish, human signal)
  • W1-C adaptive tiers (user 2026-10-09, contract 2026-10-09-model-router-tiers-1237, plan r4 after 3 lenses + 2 confirmations: 6 BLOCKER + 29 MAJOR closed by named changes): phases name absolute tiers (best fable>opus>sonnet · big opus>fable>sonnet · work sonnet>opus · cheap haiku>sonnet); breaker fed by classic.StopFailure kinds rate_limit|overloaded|billing_error|model_not_found + PostModelSwitch auto (episode backoff 15→300 min, /model clears, /clear keeps, /route reload clears); fallback chain fable→opus→sonnet→haiku; main UPGRADE by default under upgradeMaxTokens 200k (fails closed on unknown usage), DOWNGRADE gated by mainModelSwitch; sticky turnModel per turn; derived orchestrate on background dispatches; prompt default rules (plan/reflect FR+EN, Unicode guards, skipped on /…, floor match, mid-turn). 58 tests; verifier CONFORME then 3 gap/hardening rounds; security PASS
  • W1-C live verification part 1 (2026-10-09 after /reload-plugins, skills-dir copy loaded, 12 hooks): derived orchestrate on a real background Explore (main high → medium while it ran → high after its end); Explore on sonnet/medium; route plan → next request xhigh (engine record); Skill(effort-low) bridged → next request low (engine record); /route show resolves all 10 phases to full ids, down: none; steps arrive as bare claude-fable-5-1 (no [1m]) so the suffix carry never fires on this session
  • W2-A mod (2026-10-09, bb56f3e on feature/model-router-w2, contract 2026-10-09-model-router-w2a-1546, plan r4 2026-10-09-model-router-w2-1546): user decisions — effort-* skills deleted (W2-B), pins REWORKED not deleted (rows = phases by role; frontmatter model:/effort: kept as census-locked off-state floor after a robustness BLOCKER), orchestrators declare phases, slim model gate. Phases write work/high + apply work/low; 56 skill rows, 21 agent rows; agents' model at spawn within tier, upward only, project-defined agents skipped (agent.offer); typed slash → name-bound marker (composer|sdk|bridge) + pending slot + idle fallback; best-tier rows in a runMain slot surviving turn end; unrowed skill leaves the route; route answer always names the id; null override rows. 3 lenses + 2 confirmations (1 BLOCKER each round closed), feater + 4 gap rounds, GATE 0 MET, verifier 3× ECARTS on coverage clauses only → user accepted at the cap, security PASS (2 MEDIUM pre-existing: first-load kill switch fails open, ReDoS size-bounded). Kit 58 → 88 tests. Doc-sync skipped for A (mod-only, docs at B).
  • W2 gate A→B (2026-10-10, engine records): typed /status → main low then medium during the dispatch; analyzer opus/xhigh from step 0 (frontmatter high) = spawn bookkeeping first, row beats frontmatter; status-reporter haiku step 0. Probe 3 dropped (gate always calls route). Probe 4 (typed skill while a background agent is live: marker vs fallback) NOT observed → limit in lib/effort-shift.md; re-check in the records whenever the user types a skill with an agent alive.
  • W2-B repo migration (2026-10-10, 1f2d33b, contract 2026-10-10-model-router-w2b-1045): 15 citers → mcp__model-router__route per phase, effort= on built-in judgment dispatches + SDD sonnet implementers (effort="medium"), lib/effort-shift.md + lib/model-gate.md rewritten (witness = route answer, /route on), deletions (effort-, effort-pins., model-check.sh, tests, installer blocks), bridge removed from the mod, census rewritten (140 checks, drift lock rows ↔ frontmatter), analyzer xhigh, CLAUDE.global.md Design-work line. Verifier ECARTS(7) → CONFORME 10/10, security PASS, full make test green (design-tool-gate env red). Docs P1-P8 user-approved; SemVer: breaking → next release 3.0.0 (MIGRATION section at release).
  • W2-A residuals CLOSED 2026-10-11 (first load fail closed, nested quantifiers, errWord, alias keys, offers capped at 256, __proto__/constructor keys tested). Known limits (unchanged): skillCalls/spawning counters are global; offers keyed by name only; builtin /effort is not a lever inside a run.

2026-10-10 — model-router wave 3-A: first-use route confirmation + decision memory (feature/model-router-confirm)

Plan .claude/tasks/plans/2026-10-10-model-router-w3a-confirm-1201.md r4, contract 2026-10-10-model-router-w3a-confirm-1201. User decisions 2026-10-10: tracked routing.json reached through the plugin dir (no new link); blocking $.ui.ask dialog at first use; per skill row, agent row, main phase.

  • W3-A landed (22455c0, 2026-10-11): routing.json = phases + rows + decisions; T1/T2/T3 dialogs; project exceptions projects[<normalized remote>] in the tracked file (project-tree layer dropped: security); writers = dialog + /route ask; /route pending; census from the file with the changed WARN exemption; kit 86 → 190. Gates: 3 lenses + 1 confirmation (BLOCKER census), feater + 4 rounds, verifier CONFORME + re-verify, security BLOCK(1) credential leak fixed → PASS, full make test green (design-tool-gate env red). Live: T2 dialogs answered by the user (verifier Keep, feater Keep, security-auditor → implement then reset to verify on user go).
  • W3-B dialog with context (2026-10-11, 604a6c4, contract 2026-10-11-model-router-w3b-dialog-1240, plan r3): descriptions (skill frontmatter / agent.offer / phase about), real id + effort in the question, Later / Keep / Change, model → effort (derived from the phases of that model) → scope, rows stay phases, T3 Later/Keep. 3 lenses (2 BLOCKERs on my r1 avoided) + 1 confirmation, feater + 2 rounds, verifier CONFORME (3rd), security PASS (4 LOW: symlinked SKILL.md read, TOCTOU, raw key in a log, any plugin hooking AskUserQuestion could answer). Live decisions: security-auditor → judge (kept, floor aligned model: opus), plan-challenger → verify (reset on user go), orchestrate Keep (T3 live check done).
  • "auto mode: use Bash/sed" block attributed to the mod: traced 2026-10-11 by reading, it is the harness's own auto-mode instruction rendered after the mod's MCP block; not in the mod source. Closed.
  • W3-C /route forget (2026-10-11, cd8d72f, contract 2026-10-11-model-router-w3c-forget-1457, plan r3): name/all/projects, guarded restore, confirmation dialog for all/projects, no-write paths, docs. 3 lenses + 1 confirmation, feater + 3 rounds (1 real defect: ForgetPlan typo), verifier at the cap on coverage (user accepted), security PASS. Kit 232 → 284. Live: orchestrate + escalate T3 Keeps seen.
  • W3-C residuals closed in W4-A (68d7ea6). Live smoke of /route forget from the terminal still unseen.
  • W4-A transition logs + residuals (2026-10-11, 68d7ea6, contract 2026-10-11-model-router-w4a-logs-residuals-1637, plan r3): verbose = transitions only (main line, hand-off per dispatch, agent moves, answered-by anomaly), stream lines removed, host prefix respected; residuals of W3-A/B/C closed (a)-(g). 3 lenses (BLOCKER: agent moves decided at spawn) + 1 confirmation, feater + 4 rounds, verifier at the cap on coverage (user accepted), security PASS + 2 MEDIUM fixed. Kit 284 → 358.
  • W4-A residuals CLOSED 2026-10-11: errWord; logCall → tryLog, noteMain refresh under safely.
  • push-guard argv check (2026-10-11, bugfix/push-guard-argv): the text match is now a prefilter; the command is split into simple commands (awk tokenizer: quotes, escapes, separators, heredoc bodies) and a push must be git's own subcommand (push/send-pack; subtree, wrappers, assignments, no-awk keep the text verdict). grep/heredoc/git log -S/git stash push over-blocks closed; "git" push miss closed. Suite 98 → 124 checks.
  • W4-B classifier (2026-10-11, e54a2b5, contract 2026-10-11-model-router-w4b-classifier-1858, plan r3): $.model.classify on plain composer prompts with no route in force → turn effort only (design→plan, analysis→reflect, coding→implement, bookkeeping→apply, chat→nothing), 3-s race, abort not counted, 3 failures pause cooldownMinutes, no dialog, no file write, classify config key, /route classify on|off, status + verbose line. 3 lenses (BLOCKER: unbounded await) + 1 confirmation, feater + 1 round, verifier CONFORME (2nd), security PASS. Kit 358 → 428. Deviation from the TODO line: no first-use confirmation on classified phases (noise per prompt; T3 already confirms the phase).

2026-09-30 — Higgsfield pack: CLI + skills in the install process, off by default (feature/higgsfield-pack)

Contract .claude/tasks/contracts/2026-09-30-higgsfield-pack-1412.md, spec + plan under docs/superpowers/ (transient). Approved 2026-09-30: toggle pack off by default, two toggles, routing lines, complete scope, TTY fix on ctx7 + 21st (option A), deny aliases.

  • /ship-feature run: 9 plan tasks, fix wave after the final review, doc sync, registries (BDR-109)
  • user decision: close the remaining npm global-install spellings in settings.json deny (npm i <pkg> -g, npm add -g, npm -g i) — pattern grammar for a mid-string wildcard unverified (BLK-025) — CLOSED 2026-10-06 (reconcile, user go: soft_deny settings.json:472 names npm add -g + flag-after-package)
  • feature/higgsfield-pack merged into develop 2026-09-30 (df6dbce, user go)
  • soft_deny "Global npm installs" entry merged 2026-09-30 (chore/npm-global-soft-deny)
  • (was) user decision pending: npm global installs from deny to a prompt tier — ask does not prompt under defaultMode: auto (BDR-090); option = one autoMode.soft_deny entry (vet the package first), deny lines removed by the user by hand — DONE 2026-09-30 (soft_deny "Global npm installs", 95168c1; reconcile 2026-10-06)

2026-09-27 — case 4 of the 6-repo review: reticle parked with a pilot recipe (chore/six-repo-review-notes)

User go "parquer avec la recette". Real gap (runtime store state, structured verdicts with file:line, replayable flows, CI gate --since), no current project needs it; cost = per-project build instrumentation in the client repo, ~4.9k tokens of MCP schemas per session, PostHog telemetry on by default, a skill that auto-runs init against "ask, don't guess". Trigger: first app-type project (client state, forms, auth, cart).

  • P1 external opt-in in lib/toggle-external.sh + profiles (off by default, qa-class); @reticlehq/server pinned in plugins.lock.json (plugin v3.3.0 seen 2026-09-27), never @latest.
  • P2 local wrapper skill replacing theirs: npx @reticlehq/server init --dry-run shown to the user, never run by the agent (runbook doctrine); MCP env DO_NOT_TRACK=1 RETICLE_TELEMETRY=0; verify only after the user ran init.
  • P3 pilot on staging only (secret redaction is name-based); flows committed, evidence gitignored; gate --since in CI evaluated before adoption.
  • P4 measure tokens per verification loop vs gstack browse on one page. Sources read 2026-09-27: docs/what-is-recorded.md, telemetry.md, token-efficiency.md, enterprise.md (core verification free; SSO/SCIM/RBAC/ policy gates under ee/), LICENSE split FSL-1.1-ALv2 server+init, Apache adapters/core/engine. Stars 898, created 2026-06-11, 2 551 commits.