daily-fix: research-pm.md over 47000-byte ratchet cap, main

Overview / Motivation

Auto-filed by the /daily 2026-07-22 parked-candidate routing pass (Step C) from a recursion-guard-parked prose follow-up on task #1611 (emitting agent: implementer report). FLEET-BLOCKING: the ratchet test is red on pristine origin/main.

Goal

Bring .claude/agents/research-pm.md back under its 47,000-byte workflow_lint.py grandfather ratchet cap (per the #829 relocate-to-rules pattern), or deliberately raise the cap with a recorded justification, so tests/test_workflow_lint.py passes on pristine main again.

Workflow gap

• Bug observed: research-pm.md regrew to 47,861 bytes > its 47,000-byte workflow_lint.py grandfather ratchet cap, failing tests/test_workflow_lint.py (1 failed / 527 passed per the #1611 implementer report) on pristine origin/main fleet-wide — every Step-9c gate that selects that test now sees a pre-existing red.
• Why it is a workflow gap: the ratchet exists to force deliberate size decisions on always-loaded agent files; an over-cap file left red poisons the shared test oracle for every concurrent session.
• Confidence (emitter): high
• verified-at-filing: stat -c %s .claude/agents/research-pm.md → 47861 bytes; grep -n 'research-pm' scripts/workflow_lint.py → cap entry "research-pm.md": 47_000 at scripts/workflow_lint.py:11165 (presence claim, both targets bind), 2026-07-23 UTC.

Proposed change (candidate diff sketch — refine in planning)

Trim research-pm.md back under the cap by relocating per-scenario content to .claude/rules/ (the #829 pattern), OR raise the cap with a Ratchet budget: justification line. The planner decides which; the trim is preferred (the cap exists for context-load reasons).

Scope / surfaces

• Primary target: .claude/agents/research-pm.md
• Secondary (only if the raise path is chosen): scripts/workflow_lint.py cap table + its test.

Constraints / invariants

• Workflow-surface only. uv run pytest tests/test_workflow_lint.py green after the change; no PM behavioral content silently dropped (relocated content stays reachable via .claude/rules/ pointers).
• This session runs under a workflow_fix_target: Provenance line — it MUST NOT auto-route any of its own subagents' workflow-fix candidates (recursion guard).

Provenance

• fingerprint: eead9cd66a72

• workflow_fix_target: .claude/agents/research-pm.md

Verbatim parked candidate (prose park, task #1611 events 2026-07-23T04:37:15Z):

parked — running under EPM_WORKFLOW_FIX_SESSION / workflow_fix_target, see workflow-fix-on-bug.md § Recursion guard. source: prose-followup (implementer report). target_file: .claude/agents/research-pm.md. bug_observed: research-pm.md regrew to 47861 bytes > its 47000-byte workflow_lint.py grandfather ratchet cap, failing tests/test_workflow_lint.py (1 failed / 527 passed) on pristine origin/main fleet-wide. proposed_change: trim research-pm.md back under its ratchet cap per the #829 pattern (relocate per-scenario content to .claude/rules/), or deliberately raise the cap with a 'Ratchet budget:' justification. confidence: high. related_task: #1611.