The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions in SKILL.md define a workflow that creates an indirect prompt injection attack surface. By requiring the agent to validate documentation by executing embedded code examples, it may execute malicious code provided by an untrusted user. \n
Ingestion points: User-provided source code, docstrings, and API definitions intended for documentation or validation (SKILL.md). \n
Boundary markers: Absent; there are no instructions to use a sandbox or to sanitize documentation examples before execution. \n
Capability inventory: Execution of shell commands including python -m doctest, pytest, tsc, and npx (SKILL.md). \n
Sanitization: Absent; the agent is directed to iterate and fix examples until they pass validation, which could facilitate the execution of complex payloads. \n- [COMMAND_EXECUTION]: The core workflow requires the agent to execute several shell commands such as python -m doctest, pytest --doctest-modules, and tsc --noEmit. These are standard tools for validating documentation and code correctness but represent a functional risk when processing untrusted input. \n- [EXTERNAL_DOWNLOADS]: The skill references external documentation hosted on GitHub Pages (jeffallan.github.io) and recommends installing various documentation utilities from official registries, such as eslint-plugin-jsdoc, pydocstyle, and interrogate.

code-documenter