Delete the redundant focused-shard-path test and correct the stale claim that the focused runner refuses every skip

[prflow-implementer]

Summary

• Deletes the slowest single test in the suite — the focused-shard-path method in lib/test/test_module_runner.py that drove three real modules serially (~151s, ~40% of its file's wall clock) — because the guarantee it made is already produced on every CI run by the shard partition (modules-large/modules-rest).
• Corrects eight stale statements that claimed the focused runner refuses every skip; since issue Route the three chmod-000 unreadable-file arms in review-stall-backstop.sh through the sanctioned skip helper #838/Give the module tier a skip channel: host-capability skips become fatal shard failures on the #877 shard path #887 the focused runner folds a sanctioned module_host_capability_skip declaration into a visible skip, while a raw skip call stays a fatal contract violation.

Changes

Test suite (lib/test/test_module_runner.py): Removes test_the_three_declaring_modules_run_green_through_the_focused_shard_path and its docstring; corrects the neighbouring exact-floor test's docstring, which had claimed a focused run "dies at the first" host-capability arm.
Project memory (CLAUDE.md): Rewrites the second ground of the full-suite-fallback set so it rests on the corrected rule (raw helper barred; host-capability declaration folded) rather than the stale claim; the set still has four grounds with the first, third and fourth unchanged.
Contributor guide (CONTRIBUTING.md): States the two-part rule once, carrying the raw-helper half and pointing at the existing module-authoring gloss for the declaration half.
Internal docs: Corrects the self-skip and tier-asymmetry passages in docs/internal/DEVFLOW_SYSTEM_OVERVIEW.md, plus one statement each in docs/internal/operations/development.md, docs/internal/claude-md-tiered-suite-rationale.md, and docs/internal/implement-skill.md; refreshes the verified-against provenance markers on the two pages that carry one.

Resolves

Resolves #2132

Test Plan

• The pooled Python suite still passes with one fewer test: lib/test/run-shard.sh python-pool reports 6434 passed / 0 failed / 0 skipped, identical to the pre-edit base aggregate, with no failure identifier naming lib/test/test_module_runner.py.
• The internal-docs structure lint is clean: lib/test/run-shard.sh monolith reports 0 failed with no new internal-docs structure violation in its retained log.
• grep -c "def test_" on lib/test/test_module_runner.py reports 115 (down from 116), and git grep -n "the_three_declaring_modules" returns no output.
• Test authoring waived: no automated test is added. This change deletes one test and corrects prose; the deleted method's coverage is retained by the shard partition, which drives the same entry point for all three modules on every CI run. Prose accuracy is not a runtime behaviour, so no test asserts the corrected sentences.

Post-Merge Verification

The following items can only be verified after this PR is merged or deployed. Tick each after performing the check.

• lib + python tests and lint (shellcheck + actionlint + ruff) are green on the merge commit.

Visual Changes

N/A

Breaking Changes

None

Prompt-surface growth

prompt-surface growth: no tracked *.md under skills/, agents/, or .prflow/prompt-extensions/ changed between 352dc01ccef94d5ded525d36f4ce20c4107ca682 and 152666f82ed1da79abe7db5d2ed73e084310511d — no table rendered.

Generated via /prflow:implement (v2.36.15, claude-opus-4-8, low)

[prflow-implementer]

/prflow:review

[github-actions]

PRFlow Review — PR #2138

Status: ✅ APPROVE
Diff profile: engine_self_modifying (full checklist); not config_only, not small_diff; no new types; no detect_all_audit
Run: View run
Reviewed HEAD: 152666f
Last updated: Phase 4 complete

Blueprint

• Phase 0.5 — diff classification
• Phase 1/1.5 — checklist generation (14 items)
• Phase 2 — checklist verification (14 passed)
• Phase 3 — review agents (5 dispatched, 0 findings)
• Phase 4 — aggregation & verdict
• terminal completion

Verdict: APPROVE (redundant-test deletion + accurate stale-prose corrections; all 14 checklist items pass, no review findings, CI green)

Issue Compliance

Reviewed against issue #2132: Delete the redundant focused-shard-path test and correct the stale claim that the focused runner refuses every skip — criteria from workpad; scope unchanged. Requirement-based checklist items are included in the verification results below.
Criteria came from the /prflow:implement workpad comment — this run's authoritative set, possibly narrowed from the issue.
No divergence between the workpad's criteria and the issue body's.

Verification Checklist Results

14 passed, 0 failed, 0 inconclusive — 0 via lite probe, 14 via agent.

✅ Passed items (14 of 14) — click to expand

• VC-1: test_module_runner.py has 115 test-method definitions (down from 116); the_three_declaring_modules absent tree-wide at head [lib/test/test_module_runner.py]
• VC-2: python-pool shard records no failure attributable to the deletion (no dangling reference; CI python-pool: success) [lib/test/test_module_runner.py]
• VC-3: coverage-map.json, registry minimum_assertions, and run.sh shard-union assertion unchanged (absent from diff) [lib/test/test_module_runner.py]
• VC-4: no import/constant/helper orphaned by the deletion [lib/test/test_module_runner.py]
• VC-5: exactly eight statements corrected across seven files, each stating the corrected rule [docs/internal/DEVFLOW_SYSTEM_OVERVIEW.md]
• VC-6: two corrected statements (DEVFLOW tier-asymmetry paragraph; CLAUDE.md fallback ground b) use distinct phrasing [CLAUDE.md]
• VC-7: CLAUDE.md full-suite-fallback set still four grounds, only (b) changed [CLAUDE.md]
• VC-8: CONTRIBUTING.md checklist step/gloss unchanged; corrected statement points at the gloss [CONTRIBUTING.md]
• VC-9: may not self-skip returns no occurrence outside lib/test/ at head [docs/internal/DEVFLOW_SYSTEM_OVERVIEW.md]
• VC-10: development.md still carries its five contract headings [docs/internal/operations/development.md]
• VC-11: provenance markers carry re-read revision 01daaee and its author date 2026-08-29 [docs/internal/implement-skill.md]
• VC-12: follow-up issue Follow-up to #2132: exact-floor policy for regenerate-artifacts and correct the lib/test/modules/ self-skip header comments #2134 exists recording the two open gaps, linked from Delete the redundant focused-shard-path test and correct the stale claim that the focused runner refuses every skip #2132 [lib/test/test_module_runner.py]
• VC-13: corrected docstring accurately describes run-module.sh host-capability fold behavior [lib/test/test_module_runner.py]
• VC-14: raw skip fatal on both tiers; wrapper folded on both tiers (matches run-module.sh + module-harness.sh) [docs/internal/DEVFLOW_SYSTEM_OVERVIEW.md]

Code Review Findings

No findings from any of the five Phase-3 agents (code-reviewer, silent-failure-hunter, comment-analyzer, pr-test-analyzer, requesting-code-review).

over-grade annotation: no finding flagged
truthfulness sweep: no finding promoted
intra-diff contradiction scan: no contradiction found

Verdict Criteria

• No FAIL / INCONCLUSIVE checklist items → no REJECT.
• No review-agent findings at or above the critical threshold → no REJECT.
• No self-contradicting-diff findings — all corrected prose was verified TRUE against lib/test/run-module.sh (issue Give the module tier a skip channel: host-capability skips become fatal shard failures on the #877 shard path #887 fold; raw skip fatal) at head.
• Full checklist ran (engine_self_modifying); 0 agents failed.
• Result: APPROVE.

Notes

• CI ground truth for reviewed commit 152666f (from the engine block): lib + python tests, all four shards (monolith / modules-rest / modules-pin / modules-large / python-pool), and lint (shellcheck + actionlint + ruff) all report success — this is the authoritative test evidence for this commit and corroborates AC "python-pool records no failure attributable to the deletion".
• The deleted test's coverage (three declaring modules shard-compatible / run green) is produced on every CI run by the modules-large/modules-rest shards, which drive the same focused-runner entry point; the host-capability fold behavior remains covered by the retained test_focused_skip_flows_through_the_shard_tally_as_a_skip.
• Prompt-surface writing-skills routing gate: no trigger-glob path touched (only CLAUDE.md, CONTRIBUTING.md, docs/internal/, lib/test/) — gate is a no-op for this diff.
• Prompt extension resolved status: content-present (this repo's review extension); final-pass requesting-code-review extension loaded-empty, resolved-root matched.

Run telemetry

Per-phase call/token/wall-clock figures: unavailable (not instrumented on this tier). Durable --persist record is gated to writable runs; this is a read-only cloud review, so the record is not persisted to the telemetry branch.

Subagent effectiveness trace

Iteration 1

• Diff profile: engine_self_modifying
• Phase 3 agents dispatched: 5
• Checklist verifiers: 0 lite, 14 agent.
• Effectiveness signal: verdict contribution (standalone review applies no fixes) — 0 of 5 agent(s) contributed (a clean APPROVE with no findings; every agent corroborated the no-defect conclusion).

[prflow-reviewer]

Verdict: APPROVE — full report in PR comment

The complete review report (checklist results, findings, details) is in the
PRFlow Review progress comment on this PR.