The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill operates by analyzing repository source code and has the authority to modify test files, creating a surface for potential indirect prompt injection.\n- Ingestion points: The skill ingests existing code and test files to perform activities like planning and reviews, potentially processing untrusted data if the repository contains external contributions.\n- Boundary markers: The instructions do not mandate the use of explicit delimiters or boundary markers when the agent processes or outputs external code content.\n- Capability inventory: The skill is authorized to perform filesystem write operations to create or improve automated tests.\n- Sanitization: While the skill contains rules to preserve unrelated code and only modify production code upon request, it lacks explicit sanitization or validation protocols for the data it ingests.

awesome-tests