Use when acting as the adversarial spec reviewer — raises steel-manned objections across six categories before plan approval, requires evidence per objection, and discloses what was not challenged
Use when designing, reviewing, or classifying an agent that guards the human — defines the sentinel category, the three-part signature (S1 read-only, S2 advisory-to-human, S3 explicit honesty rule), the near-miss gallery that explains why…
This skill should be used when the user asks to "assess AI literacy", "run an assessment", "check literacy level", "evaluate our AI collaboration", "where are we on the framework", or wants to determine their team's AI literacy level using the ALCI instrument.
Use when running the shared drift-detection logic that backs /harness-audit and /harness-sync — produces a structured drift report covering convention files, ONBOARDING.md, snapshot staleness, constraint regressions, recurring reflection patterns, and…
Use when checking harness health, setting up observability cadences, understanding snapshot formats, configuring telemetry export, or verifying that the harness's own observability is working — covers all four layers of harness observability
Use when acting as the decision-archaeology agent — surfaces decisions a spec has made (including the silent ones), emits each material choice as a Henney-style pattern story for human disposition, and pays down intent debt before plan approval
Use when mapping the voices a spec affects and drafting the concrete question worth asking each one — the charter, the line the agent never crosses, where voices come from, what makes a question worth asking rather than a meeting worth scheduling, the…
Use when closing an agentic session deliberately rather than by attrition — carries the four-step ritual (survey, park, closure summary and reflection, close), the anchor grammar behind the next-action question, the evidence for why a specific written plan…