The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze and summarize external "source repos" provided by the user (as described in Phase 0 and goal.md). This ingestion of untrusted data creates a surface where malicious instructions embedded in the code, comments, or tests of the analyzed project could attempt to influence the agent's behavior or output. The skill mitigates this through the use of strict, predefined templates and a requirement to treat the source material as a read-only reference.
Ingestion points: The agent reads interaction state, domain objects, and behavioral tests from an external source repository (documented in SKILL.md and goal.md).
Boundary markers: The skill uses specific Markdown templates (document-template.md, goal-template.md) and a glossary to maintain output consistency.
Capability inventory: The agent has the capability to write files to a new local directory and execute the included check-links.py script.
Sanitization: No explicit sanitization or filtering logic is defined for the content ingested from the source repository.
[COMMAND_EXECUTION]: The skill instructions in SKILL.md and goal-template.md direct the agent to execute a local Python script (check-links.py) to verify links and anchors. The agent also uses standard shell commands like git init and git rev-parse. The Python script is included with the skill and uses standard library modules to perform regex-based text parsing and filesystem traversals.

product-description