The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from external sources such as git history and workspace evidence files during its learning loops.
Ingestion points: The skill reads git commit history and issue trackers for Change Decision Records (ChDRs) and scans session state/evidence files (SKILL.md).
Boundary markers: The orchestrator utilizes a hybrid supervision default with mandatory human gates enforced at the clarify step and during final PR creation.
Capability inventory: Coordinates internal skills via the shared executor, interacts with the local file system in the .adlc/ directory, and manages repository PRs.
Sanitization: Implements interactive provenance reviews and approval gates to validate mined data before it is promoted to centralized directives.

factory-learn