mind: record feature-ranker-ignores-header-keys shipped - #176
Merged
Conversation
The ranker fix shipped as PyAutoBrain#217 (e099383, 291 passed / 0 failed on 3.12 + 3.13). Records it to complete/2026/08/ and removes the draft. The record keeps the three things worth not rediscovering: - the fenced-block skip is load-bearing, not tidiness. The bug prompt QUOTES the offending header in a ```-block, so a naive line scan makes the bug report declare itself blocked. lifecycle.py:draft_gate_refs had already hit this and solved it the same way. - length is a bad size proxy: the derived score weights prompt size, and a prompt grows as it accumulates findings, so a long well-documented prompt derives too-large for work that is not. That is what a declared Difficulty: is for. - test_skill_install.py fails 2 tests in a cloud session and passes in CI (the installer reports "web-github / ci-only"). Do not chase them locally. Also records a SECOND instance of the same defect class, found while doing this and deliberately not fixed here: `pyauto-brain bug` on this very prompt answered "looks like a feature/ task, not a bug" — because the prompt says "Feature Agent" repeatedly and the Bug Agent classifies on prose keywords while ignoring the declared `Type: bug` header. Same family, different conductor, currently unfiled. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01WSCA1h6cSKMNwFmWEsxuCL
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The ranker fix shipped as PyAutoBrain#217 (
e099383; pytest 3.12 + 3.13, 291 passed / 0 failed). Records it tocomplete/2026/08/feature-ranker-ignores-header-keys.mdand removes the draft.lifecycle.py checkandindex --checkboth OK.What the ranker now does
_feature.pymentionedDifficultyexactly once — where it printed its own re-derived score — and read no header key at all, despiteREFERENCE.mdpromising that the value Intake persists is "the one the Feature Agent later acts on". It now honoursDifficulty:(declared overrides derived, disagreement reported),Priority:(orders the shortlist), andStatus: blocked/Blocked-by:(never the recommended pick).Measured on this repo's live backlog: the offending prompt drops rank 1 → last of 27, two further blocked prompts surfaced that nobody had spotted by hand, and
draft_staleness_detection_signalsmovestoo-large#26 →medium#5.Three things the record keeps
lifecycle.py:draft_gate_refshad already hit this and solved it the same way.too-largefor work that is not. That is what a declaredDifficulty:is for.test_skill_install.pyfails 2 tests in a cloud session and passes in CI (the installer reportsEnvironment: web-github / ci-only). Recorded so nobody chases them locally.A second instance of the same defect class — recorded, not fixed
Running
pyauto-brain bugon this very prompt returned "Fix strategy: defer/re-home (looks like a feature/ task, not a bug)" — because the prompt says "Feature Agent" repeatedly and the Bug Agent classifies on prose keywords while ignoring the declaredType: bugheader. Same family as the bug just fixed, different conductor. Left unfixed and unfiled deliberately: it is its own task, and folding it in here would have made a small fix a speculative one.🤖 Generated with Claude Code
https://claude.ai/code/session_01WSCA1h6cSKMNwFmWEsxuCL
Generated by Claude Code