The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill is designed to ingest and act upon data from external session reports, which could potentially contain instructions intended to influence the agent's behavior.
Ingestion points: Data is extracted from codagent:session-report files during the 'Extracting findings' phase.
Boundary markers: The skill does not define specific delimiters or 'ignore' instructions for the content being extracted from reports, though it does use a subagent to help isolate the extraction process.
Capability inventory: The agent is instructed to perform code edits, execute commits, and dispatch subagents based on the findings in the reports.
Sanitization: No explicit sanitization or escaping of the report content is mentioned, although the skill requires 'spot-checks' and evidence citation to verify claims made in the reports before acting.

review-assumptions