The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill is designed to process external security findings and repository content, which could potentially contain adversarial instructions. While the skill lacks explicit input delimiters, it mitigates this risk through a robust multi-agent review process.
Ingestion points: The skill receives external data via the 'finding' description and the 'repository root' context (SKILL.md).
Boundary markers: There are no specific instructions to wrap external findings in delimiters or use explicit ignore-markers.
Capability inventory: The skill enables the agent to perform file modifications and execute shell commands for builds and tests (SKILL.md).
Sanitization: The workflow incorporates a 'Bypass and regression reviewer' agent to validate changes, serving as a procedural control rather than technical input sanitization.

fix-finding