feature: honour the prompt header keys the ranker is documented to act on

[Jammy2211]

PyAutoMind REFERENCE.md states the contract:

The Intake (Conception) Agent writes these automatically … sourcing Difficulty: from the shared sizing faculty the Feature Agent also uses — so the value shown up front is the one the Feature Agent later acts on.

It was not acted on. _feature.py mentioned Difficulty exactly once, where it printed its own re-derived score, and read no header key at all — not Difficulty:, not Status:, not Priority:, not Blocked-by: / Closes-when:.

The visible failure: feature select recommended draft/feature/autonomy/10_scheduled_runs.md as the next task to start — a prompt declaring Status: blocked, Priority: low and an explicit Blocked-by: gate whose body reads "Do not start before the queue runner has completed several supervised interactive runs cleanly."

Prompt: PyAutoMind:draft/bug/pyautobrain/feature_ranker_ignores_header_keys.md

Where the parsing lives

In the sizing faculty, beside the derivation it overrides — so declared and derived stay in one place, and the bug / refactor conductors (which share parse_prompt) get the same reading for free.

declared_header() reads the five keys and skips fenced blocks. That skip is load-bearing, not tidiness: the bug prompt driving this fix quotes the offending header in a ```-block, so without it the bug report would declare itself blocked. Same rule and same reason as PyAutoMind/lifecycle.py:draft_gate_refs.

Two smaller traps handled: trailing # note comments are stripped by splitting on " #" so a Repo#123 ref survives intact (the live backlog has both on one line), and an unrecognised Difficulty: value is ignored rather than trusted.

What the ranker now does

Key	Effect
Difficulty:	Overrides the derived level; the disagreement is reported, not silently resolved — the gap is evidence about the heuristic.
Priority:	Orders the shortlist, above the difficulty term.
Status: blocked / unresolved Blocked-by:	Sinks the prompt below everything; never the recommended pick, and next_action says do not start.

Blocked prompts stay listed in their own labelled band so a human can see and override.

declared_blocked() is deliberately conservative: this agent is offline and cannot tell whether a gate has since closed, so an unresolved Blocked-by: reads as blocked and the output points at lifecycle.py issues --drafts, which does resolve refs against GitHub. Being wrongly held back is cheap and visible; being wrongly recommended is the failure this exists to stop.

Measured on the live backlog

• The offending prompt drops from rank 1 → last of 27, labelled BLOCKED — Status: blocked; priority low; declared medium vs derived small.
• Two further blocked prompts surfaced that had not been spotted by hand (ep_analytic_updates, oversampled_psf_dataset_adoption).
• draft_staleness_detection_signals moves from too-large Flatten the skill chain: short dispatchers, plan_branches folded, mandatory-load budget #26 to medium Modernise and re-home PyAuto workflow skills for PyAutoBrain #5. It derived too-large purely on file length — and length is a bad size proxy, because a prompt grows as it accumulates findings. Documented in AGENTS.md as the reason the declared key exists.
• Five declared/derived disagreements are now visible rather than silent.

Verification

289 passed (was 279; the 2 test_skill_install.py failures are pre-existing and environment-only — they do not occur in CI).

New tests are hermetic and pin the regression by the offending prompt's header shape, not the live file, so fixing the backlog cannot silently retire them. Confirmed they actually bite: reverting only _feature.py fails 7 of the 10 (the 3 survivors are pure declared_header unit tests, which is coherent). Fixtures use invented repo names, so the file adds no tenant-firewall instance facts — verified against repos_sync.py --check.

🤖 Generated with Claude Code

https://claude.ai/code/session_01WSCA1h6cSKMNwFmWEsxuCL

---

Generated by Claude Code