| name | Implementation Gap Check |
| description | Structural audit skill invoked against plans prior to execution to resolve missing edge cases. |
Gap-Check Skill
Trigger condition:
You should execute this skill prior to submitting a complex implementation_plan.md for human review, or when the user explicitly requests an audit.
Executing the Gap Check:
Evaluate the proposed logic against the foundational models described in docs/AGENT_DECISION_TREE.md:
- Security / Red-Teaming Coverage: Does the mechanism rely purely on single-turn evaluation, or does it properly branch (e.g., TAP algorithm)?
- Evaluation Metrics: Is there an Auto-CoT JudgeAgent assigned to evaluate the results? If missing, the plan has a gap!
- Defense Pipeline: Will this output synthesize into dynamic guardrails, or is the loop artificially broken?
Generate a gap analysis report in the plan's Open Questions section targeting the user.