The Agent Skills Directory

[SAFE]: No malicious patterns or security vulnerabilities were identified. The skill implements a defensive 'grill' and 'review' architecture that minimizes the risk of automated misbehavior.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests untrusted data from tickets and Pull Requests, but this is mitigated by design. Ingestion points: Processes ticket/PR metadata and repository facts in doctrine.md and reference.md. Boundary markers: Uses a mandatory 'Grill' phase with 'Locked' closure and 'Active Rules' to ensure user-verified understanding of tasks. Capability inventory: Includes terminal output recording, file system implementation, and browser evidence gathering capabilities. Sanitization: Requires human review of all implementation results, mandatory /code-review invocation, and explicit user promotion for fix remediation.

goal