Lightweight single-pass AC verification before Archive — run implementation against each AC, produce pass/fail evidence. FAIL blocks Archive and rolls back to Phase 4. Mutually exclusive with ultraqa: choose ac-verify when AC count ≤ 3 AND risk ≠ HIGH; otherwise use ultraqa. See .claude/rules/skill-precedence.md Zone C.

<HARD-GATE> Do NOT write or modify implementation code until a design has been presented, alternatives documented, and approval received. </HARD-GATE>

Meta-cognitive check on reasoning quality: prevents hallucinations, overconfidence, anchoring, and narrow thinking. TRIGGER at Propose/Review phase AFTER decision-frameworks produces options — run as a second-pass bias scan on the reasoning, not as a framework for building options. Composes with decision-frameworks (which runs first); do NOT substitute one for the other. See skill-precedence.md Zone G.

Starting-from-scratch protocol for projects with no existing codebase. Replaces the standard Explorer phase's 'infer from existing code' logic with a forward-design sequence: domain model → API contract → DB schema → package structure → scaffold. TRIGGER when input-classifier emits Scenario=Greenfield or when there is no src/ directory.

FRONT-DOOR CLASSIFIER (runs INLINE on the main agent, never dispatched). Tags raw input as one of PRD / Idea / Bug / Signal / Performance / Security / Compliance / Feedback and emits an [Intake] block telling the main agent WHERE TO ROUTE — to product-manager-expert (PRD), root-cause-debug (Bug/Signal), requirement-engineer (Idea/Feedback/Compliance/Security), or Explorer (everything else). DOES NOT translate to ACs itself — that is requirement-engineer's job. TRIGGER when input has no @shortcut and is longer than one line.

Shared OK/WARN/FAIL severity contract for every gate script under .claude/scripts/gates/. TRIGGER whenever a gate is invoked — agent MUST surface WARN/FAIL inline using the reporting protocol below. New gates MUST pass `_severity_audit.py`.

A→B system migration protocol where the core requirement is behavioral equivalence, not new features. Generates an equivalence test suite BEFORE migration begins. Migration is done when equivalence tests pass. TRIGGER when input-classifier emits Scenario=Migration or when the request is 'migrate from X to Y'.

Mandatory root-cause investigation before any fix. TRIGGER when encountering any bug, test failure, unexpected behavior, or runtime exception. MUST complete Phase 1 (Root Cause Investigation via Hierarchical Localization Protocol) before proposing fixes — never jump straight to solutions.

Generate candidate improvement plans, implement one at a time, measure score delta against the eval-harness baseline, keep the best result. Stop when acceptance criteria are met or plateau is detected.

Skill navigator. List active skills, locate archived ones, choose the right sequence per scenario.

Multi-stakeholder requirement conflict resolution protocol. Detects when two or more requirements from different stakeholders are mutually exclusive, partially overlapping, or impose contradictory constraints. Produces a structured conflict map and a resolution decision (negotiate / defer / escalate). TRIGGER when: (a) user mentions conflicting instructions or requirements from different parties (PM vs tech lead, legal vs product, frontend vs backend); (b) adversarial-review Category A returns a CRITICAL mutual-exclusion finding; (c) product-manager-expert PRD ingestion finds two requirement units that directly contradict each other.

Rulebook for WAL fragment writes (domain, api, rules, data, architecture) during Archive — keeps the LLM Wiki synced without merge conflicts. User-elected per task via h-archive Step 3b; skipped entirely in PATCH. Executed by the knowledge-extractor sub-agent by default. See .claude/rules/skill-precedence.md Zone D.