workflow-fix: wire identity+bias baseline + kNN retrieval mandate into planner/critic lenses

Overview / Motivation

Auto-filed by the workflow-fix-on-bug protocol from the orchestrator's own observation while landing the 2026-07-22 CLAUDE.md standing rule "Identity+learned-bias baseline AND kNN-retrieval metric — report BOTH for every representation mapping" (user directive, chat 2026-07-22).

Goal

Wire the identity+learned-bias baseline + kNN-retrieval mandate into planner.md §6, critic.md Statistics & Measurement lens, statistics-critic.md, and experiment-guidelines.md.

Workflow gap

• Bug observed: CLAUDE.md standing rule (2026-07-22) mandates both reads for every representation mapping but no enforcing agent file names them.
• Why it is a workflow gap: CLAUDE.md describes a rule but the implementing files (planner/critic lens specs) do not enforce it — the named "CLAUDE.md describes a rule but the implementing file doesn't enforce it" class in workflow-fix-on-bug.md.
• Confidence (emitter): high
• verified-at-filing: grep -icE 'identity.{0,30}bias|learned.bias|nearest neighbo|knn|retrieval' .claude/agents/planner.md .claude/agents/critic.md .claude/agents/statistics-critic.md .claude/rules/experiment-guidelines.md → 0 hits in ALL FOUR named targets (absence-of-mandate claim; 0-hit in-target IS the evidence); git log --oneline --since='7 days ago' -- .claude/agents/planner.md .claude/agents/critic.md .claude/agents/statistics-critic.md shows no landed fix covering this (top hits: 4539d3f859 self-count rule, 00c775fb4a verify_plan c31 — unrelated) (2026-07-22)

Proposed change (candidate diff sketch — refine in planning)

• planner.md §6 (measurement): for any fitted v_X→v_Y map, the plan names (a) the identity-family baseline incl. the learned-bias form (analysis/mapping_baselines.identity_bias_predict; dimension-mismatch stated as inapplicable) and (b) the kNN retrieval read (analysis/mapping_baselines.knn_retrieval, chance = k/n_pool stated).
• critic.md Statistics & Measurement lens + statistics-critic.md: REVISE a mapping plan that omits either read without a stated exemption.
• experiment-guidelines.md: add a numbered guideline (owner: statistics-critic) pointing at the CLAUDE.md bullet + helper.

Scope / surfaces

• Primary target: .claude/agents/planner.md, .claude/agents/critic.md, .claude/agents/statistics-critic.md, .claude/rules/experiment-guidelines.md
• Grep the workflow surface for 'mapping_baselines' before editing and keep wording consistent with the CLAUDE.md bullet (the source of truth).

Constraints / invariants

• Workflow-surface only — never experiment code, configs/, or tasks/.
• scripts/workflow_lint.py --check-asks passes; if experiment-guidelines.md changes, lens-coverage-map.md consistency (--check-lens-coverage) holds.
• This session runs under EPM_WORKFLOW_FIX_SESSION=1 and carries a workflow_fix_target: Provenance line — it MUST NOT auto-route its own subagents' workflow-fix candidates (recursion guard).

Provenance

• workflow_fix_target: .claude/agents/planner.md, .claude/agents/critic.md, .claude/agents/statistics-critic.md, .claude/rules/experiment-guidelines.md
• fingerprint: ab30643f3287

Orchestrator-observed gap (no candidate block; surfaced while applying the user-directed CLAUDE.md standing rule, chat 2026-07-22: "also add to CLAUDE.md to always run this baseline and this metric for any mapping we compute from now on").