The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted content from a source codebase, creating a surface for indirect prompt injection. Maliciously crafted code comments or data within the repository could attempt to subvert the agent's summarization logic or documentation tasks.
Ingestion points: The skill reads all files and directory structures within the provided source_root during its bootstrap, survey, and exploration phases as defined in SKILL.md.
Boundary markers: The skill lacks explicit delimiters or specific 'ignore' instructions for the agent when intake of the source code content occurs, which could lead to the agent following instructions found in the code.
Capability inventory: The agent employs file system tools for listing directories, searching file contents, and reading and writing files to the local storage.
Sanitization: There is no evidence of sanitization or validation performed on the codebase content before it is parsed and used to generate the final snapshot document.

take-snapshot