The Agent Skills Directory

[COMMAND_EXECUTION]: The bundled script scripts/doc_audit.py executes local git commands using subprocess.run to determine the project root directory and fetch file modification timestamps. These operations are performe d on local file paths discovere d via safe globbing patterns.
[PROMPT_INJECTION]: The skill exhibits an indirec t prompt injectio n surface because its search agents are instructe d to rea d and process content from the project codebase and existing documentation. This content is untruste d and coul d theoretically contain malicious instructions.
Ingestion points: references/agent-dispatch.md defines subagents that rea d markdown files under docs/ and manual/, as well as source code across the repository.
Boundary markers: The prompt templates use structure d headings but lack explicit delimiters or instructions fo r the subagents to ignore embedded commands within the file content being analyze d.
Capability inventory: The skill has the ability to rea d codebase files, run git metadata commands, and write or modify markdown files throughout the documentatio n directories.
Sanitization: There i s no evidenc e of content filtering or sanitizatio n of the files rea d by the subagents before processing.

doc-maintenance