qa: sync gate guards (root-cause taxonomy + corpus manifest) with structural_completeness + narration_no_ooc_leak

[100yenadmin]

Problem

Two QA guard tests were RED on main (the CI / qa-release-gate-tests job has been failing on main for several commits). Both stem from one root cause: the gate checks structural_completeness and narration_no_ooc_leak were added to qa/assert_behavioral.py without updating the guard metadata that's supposed to track the gate.

1. test_root_cause_analyzer.py::test_taxonomy_matches_real_gate_check_names — the root-cause taxonomy (qa/BEHAVIORAL_GATE_TAXONOMY.json) was missing both checks.
2. test_behavioral_gate_corpus.py::test_manifest_covers_every_fatal_check — the gate-corpus manifest had no fixture and no TODO entry for narration_no_ooc_leak (a FATAL check: chk(... fatal=_leak_red)). structural_completeness already had a fixture, so only the taxonomy was missing it.

Both were pre-existing on main and independent of each other; the corpus failure does not read the taxonomy at all.

Fix (guard metadata only — assert_behavioral.py untouched)

Taxonomy (qa/BEHAVIORAL_GATE_TAXONOMY.json) — added both entries:

check	category	rationale
structural_completeness	ENGINE_INVARIANT	FATAL duo-lane check that a ≥10-beat companion session engaged a core relationship/quest system. Reads engine-written state — structurally parallel to party_traveled / world_advanced_time. Locations → companion.py / server.py / questgen.py.
narration_no_ooc_leak	DM_ADHERENCE	Player-facing OOC craft-scaffolding leak gate (0 pass / 1-2 WARN / ≥3 RED). About the DM's narrated prose; remediation → FICTION-ONLY mandate in skills/dungeon-master/SKILL.md + duo DM prompt; _NARRATION_LEAK_RE as the false-positive tuning point.

Corpus manifest (qa/gate_corpus/manifest.json) — ran the canonical sync python qa/gate_corpus/builder.py, which deterministically regenerated all 18 fixtures (byte-identical, zero churn) and auto-added a TODO__narration_no_ooc_leak entry (a faithful minimal OOC-leak fixture is deferred work, flagged with a reason).

Test

$ uv run --directory servers/engine python -m pytest \
    ../../qa/test_behavioral_gate_corpus.py ../../qa/test_root_cause_analyzer.py -q -p no:xdist
38 passed, 1 skipped

No behavior change to any gate; only the analyzer's RED→actionable-location mapping and the corpus coverage audit gain the two checks.

Summary by CodeRabbit

• Tests
  • Added new automated QA checks to enforce narrative consistency (preventing out-of-character information leakage) and structural completeness.
  • Introduced a new regression-corpus case for the narrative consistency check, currently marked as pending until the minimal required fixture content is available.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: c27ba28e-e8b6-49c5-b4d6-67342e93f1e0

📥 Commits

Reviewing files that changed from the base of the PR and between abb0df1 and a921a19.

📒 Files selected for processing (1)

• qa/BEHAVIORAL_GATE_TAXONOMY.json

🚧 Files skipped from review as they are similar to previous changes (1)

• qa/BEHAVIORAL_GATE_TAXONOMY.json

---

📝 Walkthrough

Walkthrough

Two new check entries are appended to qa/BEHAVIORAL_GATE_TAXONOMY.json: narration_no_ooc_leak mapped to the DM_ADHERENCE category and structural_completeness mapped to ENGINE_INVARIANT. Each entry includes candidate code locations, a duo retest command, and a descriptive hint. A corresponding todo corpus case is added to qa/gate_corpus/manifest.json for the narration check, flagged pending fixture construction.

Changes

Behavioral Gate Taxonomy Additions

Layer / File(s)	Summary
New taxonomy check entries qa/BEHAVIORAL_GATE_TAXONOMY.json	Adds narration_no_ooc_leak (DM_ADHERENCE) with candidate locations in DM skill doc, DM duo play script, and qa/assert_behavioral.py, and structural_completeness (ENGINE_INVARIANT) with candidate locations in companion, server, and questgen code paths; both entries include exact duo retest commands and threshold hints.
Corpus test case registry qa/gate_corpus/manifest.json	Adds a TODO__narration_no_ooc_leak case entry marked todo: true with a reason noting that the minimal fixture for the narration check has not yet been constructed.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~4 minutes

Possibly related PRs

• electricsheephq/WorldOS#972: Implements the narration_no_ooc_leak behavioral gate and its tests in qa/assert_behavioral.py, directly corresponding to the taxonomy entry added in this PR.

Poem

🐇 Two new gates now guard the realm,
No out-of-character ghosts at the helm!
The DM speaks only in-fiction true,
Quests and companions complete their due.
Hoppity-hop, the taxonomy grew! 🗂️

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately and concisely describes the main change: syncing gate metadata (taxonomy and corpus) with two new gate checks. It is specific, clear, and directly reflects the changeset.
Description check	✅ Passed	The description comprehensively covers the problem, fix, and validation. It includes problem statement, detailed fix explanation with a table, test results, and explicitly states no behavior changes. The Licensing/CLA section is missing.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}