项目文件夹

文件

T

Asim Aslam 85374c6401 loop: cut planner bookkeeping churn + cap diminishing-returns conformance work (#4731 )

An assessment of the last 100 commits found ~45% were pure "refresh planner
priorities" bookkeeping and much of the rest was thrashing on one weak provider
(AtlasCloud text-tool-call repair) and guarding docs the loop already wrote —
motion, not progress. Two prompt-policy fixes:

PLANNER (planner.md):
- Default to NOT committing. Post the assessment and close the issue; open a
  PRIORITIES.md PR ONLY when the change is MATERIAL (top item changes, an item
  is added/removed, or a top item's issue closed). No PRs for reorders below
  the top, reword, or "keep it current" — that churn was the loop's #1 waste.
- Add a diminishing-returns guard: don't queue the Nth doc-guard or the Nth
  robustness workaround for an already-tolerated class; mark exhausted areas
  needs-human and rank real-headroom capability instead.

TRIAGE (triage.md):
- Cap the AtlasCloud/plan-delegate tail-chase: another instance of a class the
  agent already tolerates is NOT filed as a routine patch — comment "recurred —
  capped" and, if worth more, needs-human. Real regressions (lint/tests/
  govulncheck on master) and genuinely new defects still get filed.

Prompt-only; reversible. Steers the loop toward outcomes over busy-work.


Claude-Session: https://claude.ai/code/session_01CmdEY7pYmV5zzwCjNJ4ykL

Co-authored-by: Claude <noreply@anthropic.com>

2026-07-11 20:11:58 +01:00

1.9 KiB

原始文件 永久链接 Blame 文件历史

Triage the failed CI run at RUNURL. It may be the linter (Lint), the unit/integration tests (Run Tests), the vulnerability gate (govulncheck), or the provider-conformance harness (Harness (E2E)).

Read the logs and root-cause each distinct failure. DEDUPE hard against open AND recently-closed issues — if a failure matches an existing or recurring one, comment "recurred" on that issue rather than filing a new one.

WHAT TO FILE:

Lint, Run Tests, or govulncheck failing on master — a real regression. File a scoped issue (gh issue create --label codex --label enhancement --title "<scoped fix>" --body "<root cause, where, acceptance>") so it is fixed promptly.
A genuinely NEW, distinct provider-conformance defect — file it.

WHAT NOT TO FILE (this cap matters):

Another instance of a class the agent already tolerates — a weak provider (e.g. AtlasCloud) emitting malformed / text-rendered / partial tool calls, or another plan/delegate notify/side-effect edge case. These have been hardened repeatedly with diminishing returns. Do NOT auto-file yet another routine robustness patch. Comment "recurred — repeated class, capped" on the nearest existing issue and, if it seems genuinely worth more investment, label it needs-human for a human to decide. The loop should not keep chasing one weak provider's output shape.
Transient flakes — live-model latency, provider outages, rate limits, network timeouts with no code cause. Ignore.
Anything needing a breaking or architectural change — label needs-human and describe it.

Close this issue (gh issue close __ISSUE__) when triage is done. Open any PR yourself from the shell with gh; do not use the make_pr tool.

1.9 KiB 原始文件 永久链接 Blame 文件历史

1.9 KiB

原始文件永久链接 Blame 文件历史