Commit Graph
6 Commits
Author SHA1 Message Date
bastien 2b25cb4704 feat(lib): floor-guard, diff-scoped detector of a weakened quality bar
SUPPRESS / SKIP / DELETED_TEST / ASSERT_DROP / STUB / THRESHOLD_DOWN over
git diff <base> (untracked files included), floor-guard: allow <reason>
waiver printed as WAIVED, rc 0/2/3. Mandatory verifier STEP 3, documented
under GATE 1 of verify-secure-loop.md. Suite: 6 kinds + WAIVED + CLEAN,
flip-tested. Adapted from agent-skills constraint-driven-development.
2026-09-27 20:17:36 +02:00
bastien 27f201d4aa feat(guardrails): refusal ends the attempt; doctrine-citers census; make test suite=
Root causes of the 2026-09-24 errors turned into mechanisms (BDR-100). hard_deny 'Routing around a guardrail': a refused command is never rerun through a wrapper, alias, heredoc, Makefile target, env file, other shell or other agent; the same clause in 14 agents and in the doctrine's sub-agent rule. make test suite=<file> runs one suite hermetically so the denied env-prefix form is never needed by hand. lib/tests/doctrine-citers.test.sh: every CLAUDE.md "Section" / § Label citation across skills, agents, lib, rules and hooks must resolve to a heading or bold label (flip-tested); its first run fixed rest-api-node.md. Doctrine 'After code changes' step 4: a changed rule, heading, label or threshold → grep every citer in the same commit.
2026-09-24 20:58:25 +02:00
bastien 9da5d8d52c feat(guardrails): push every commit, static deny for destructive tools, brief carries no user authority
Layer C of the plan written after the 2026-09-21 wipe (BDR-095): a reviewer
sub-agent traced `lftp mirror --delete` against a local file:// tree, the
prose tiers named neither lftp nor a local trace, the brief had authorized
it, and four days of commits had never left the machine.

- gitflow: `start` pushes the branch with its upstream, merge targets are
  pushed after each merge, and `init`/`install-hook` write post-commit and
  post-merge hooks that push every commit as it lands (warn, never block;
  GITFLOW_NO_PUSH=1 for throwaway repos). T18 + T19 (installed == emitted).
- hooks/unpushed-guard.sh on SessionStart and Stop: branch ahead of its
  upstream, no upstream, or no origin. Non-blocking systemMessage.
- settings.json: static deny for transfer and mirror tools, rsync --delete,
  xargs rm, pipe-to-shell, chmod/chown -R, sudo/doas/pkexec, disk tools,
  chattr, docker volume drops/prune/--privileged/socket/-v /:, git history
  destruction, --no-verify and core.hooksPath; new hard_deny "destructive
  tool against a local path, brief carries no user authority"; soft_deny
  reworded + discarding uncommitted work; environment records the incident.
- CLAUDE.global.md "Destructive tools & data loss"; the four report-only
  agents trace by reading, never by running, whatever the brief says.
- lib/tests/guard-bash.test.sh: executable spec of the PreToolUse guard
  (214 cases). The hook itself is not shipped (BLK-022); the spec skips.
2026-09-22 07:43:12 +02:00
Bastien Chanot 63310467ca feat(gates): deterministic floor (GATE 0) under the fresh verifier
GATE 1 is an LLM dispatch and the verifier's mandatory PROOF: line is a line
the verifier writes — nothing structurally stops it being produced without
anything being executed. Nothing deterministic sat between the executor and
that dispatch.

An acceptance criterion can now carry an oracle: indented CHECK: (command),
EXPECT: (success-only marker), EVIDENCE: (slot). lib/gates.sh runs them
fail-closed — MET requires exit 0 AND the marker, so a nonzero process never
passes on its error text carrying the token — and writes the outcome back
into the contract, so the fresh verifier reads evidence as fact rather than
trusting the executor's report.

GATE 0 runs that floor before any verifier is dispatched; a red build sends
the executor back for free, on its own iteration budget. ABANDON: <id>
<reason> turns an impossible criterion into a visible handoff that blocks
CONFORME and routes to the human gate, via the new ABANDONED(n) verdict —
a distinct token because it routes distinctly, never a dev loop. feater and
bugfixer gain a four-pass completion discipline, scoped so a pass can never
widen the contract.

The runner's parse fails closed on partial oracles, duplicate ids,
unindented attributes and runnable criteria with no EVIDENCE: line, and
executes nothing at all when the ledger is malformed. status never executes
and never writes; run always re-executes, since trusting written evidence is
the failure being closed.

Adapted from the unlazy skill (Leonxlnx/unlazy, MIT). Its Stop hook,
approval store, .unlazy/ tree, depth-tree arithmetic and Node checker were
deliberately refused — BDR-083 records each reason.

64 assertions in lib/tests/gates.test.sh, non-execution proved by sentinel
with its own positive control asserted first.
2026-08-24 13:12:38 +02:00
Bastien Chanot ea6c126f73 job9: B pin verifier to sonnet 2026-07-08 02:50:08 +02:00
Bastien ChanotandClaude Opus 4.8 6aed5eea8c feat(agents): contract interview include + verifier agent (verify-loops lot 2)
lib/contract-interview.md: mandatory upstream passage for all orchestrators.
Verbatim REQUEST (immutable), proportional questions (complete request =
zero, max 3 one batch), testable criteria + file scope, written to disk
immediately (.claude/tasks/contracts/<date>-<slug>-<HHMM>.md — a context-only
contract dies at compaction). Lifecycle: enrichment only at human gates
([gated] marker, scope micro-gate), supersedes for re-scope, aborted runs
deleted or committed status:aborted — never left dirty. Hand-off = path,
not restatement.

agents/verifier.md: fresh read-only verifier. Reads the contract from disk,
renders VERIFY — VERDICT: CONFORME | ECARTS(n) | ERROR. Blind: never
receives iteration history. PROOF line mandatory (LRN-048), UNVERIFIABLE
never MET, mute verifier never a PASS (retry once fresh, 2nd structural
failure = human escalation). Orchestrator protocol documented in-file
(max 3 iterations, re-verify request before security).

lib/tests/contract-verifier.test.sh: 31 deterministic structure locks on
the load-bearing doctrine clauses — green, shellcheck clean.

Behavioral dogfood (2 fresh subagents on a planted fixture): gap case →
ECARTS(2) exactly as planted (NOT-MET located + out-of-scope flagged);
conform case with injected fake iteration history → CONFORME, noise
ignored, real python spot-check as evidence. Both outputs parse-clean.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XpphkdTosUzokBDNG7PToS
2026-07-03 18:50:28 +02:00