bmottin/claude_mac
1
0
Fork 0
forked from bchanot/claude
Code Pull Requests Activity

feat(global): recalibrate instruction layer for Claude 5 family

Browse Source 
- Delegation block: model-neutral when-guidance replaces the Opus 4.8
  under-delegation counter (LRN-030 trait inverted on Opus 5; Claude
  Code injects its own anti-delegation prompt there, #80988). Gates
  carve-out keeps verifier/security/challenge dispatch mandatory.
- Drop the 'staff engineer' self-check bar (Opus 5 over-verification
  trigger per official migration guide); honest-reporting steps stay.
- Deviations bullet: finish-whole-task clause (Opus 5 scope-expansion
  counter), scoped so 'gone wrong → STOP' still wins.
- Written-deliverable length rule (Opus 5 writes ~30-40% longer).
This commit is contained in:
Bastien Chanot
2026-07-30 12:58:06 +02:00
parent eab2a10cd5
commit 0f7b565bb0
1 changed files with 10 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
CLAUDE.global.md
+10 -6
View File
@@ -22,6 +22,8 @@ Apply unless repo-specific instructions override.
- Document intent, not mechanics. Use project doc style (docstring, JSDoc…). - Document intent, not mechanics. Use project doc style (docstring, JSDoc…).
- Explicit, consistent, meaningful names. Straight control flow, - Explicit, consistent, meaningful names. Straight control flow,
no hidden side effects. no hidden side effects.
- Written deliverables (docs, reports, .md): length matched to what
the task needs — no filler sections, no boilerplate summaries.
## Refactoring ## Refactoring
- Priority: safety → readability → consistency. - Priority: safety → readability → consistency.
@@ -40,11 +42,12 @@ Apply unless repo-specific instructions override.
- Confirm before implementing only when real trade-offs exist (multiple - Confirm before implementing only when real trade-offs exist (multiple
valid approaches, breaking change, destructive action) — else proceed. valid approaches, breaking change, destructive action) — else proceed.
- Minimal changes unless broader refactor requested. State trade-offs. - Minimal changes unless broader refactor requested. State trade-offs.
- Sub-agents keep main context clean — one task per sub-agent. - Sub-agents: one task per sub-agent, main context stays clean.
More compute on hard problems. Task fans out across independent Delegate genuinely independent, sizeable tracks (wide multi-file
items (many files, parallel searches, multi-point checks) → delegate exploration, parallel audits) — not work doable in a few tool
to sub-agents, don't iterate serially. Default to delegation for calls. Skill-mandated gates (fresh verifier/security/challenge)
multi-file exploration. Counters model tendency to under-delegate. always dispatch as written. Don't redo delegated work by hand —
failed gates re-dispatch fresh executors instead.
- One question upfront if needed — don't interrupt mid-task. - One question upfront if needed — don't interrupt mid-task.
*Exception: skill-mandated gates and checkpoints (orchestrator *Exception: skill-mandated gates and checkpoints (orchestrator
validation gates, approval gates, darwin checkpoints) always fire.* validation gates, approval gates, darwin checkpoints) always fire.*
@@ -53,6 +56,8 @@ Apply unless repo-specific instructions override.
- Something goes wrong → STOP, re-plan. Never push through. - Something goes wrong → STOP, re-plan. Never push through.
- Deviations: minor or clearly justified → do, explain after. - Deviations: minor or clearly justified → do, explain after.
Significant or shaky justification → ask before deviating. Significant or shaky justification → ask before deviating.
Finish the whole task: blocked on an independent sub-part → do
the rest, state what's missing. Gone WRONG → still STOP, re-plan.
- Root causes only. No temp fixes. Never assume — verify paths, APIs, - Root causes only. No temp fixes. Never assume — verify paths, APIs,
variables before use. variables before use.
@@ -77,7 +82,6 @@ Apply unless repo-specific instructions override.
2. Report what verified, what not. 2. Report what verified, what not.
3. List remaining risks, surviving deviations. 3. List remaining risks, surviving deviations.
4. Don't mark complete without proof it works. 4. Don't mark complete without proof it works.
Bar: "would staff engineer approve?"
5. Correction or notable event → capitalize to right registry 5. Correction or notable event → capitalize to right registry
(see "Memory registries"). (see "Memory registries").