The Agent Skills Directory

[COMMAND_EXECUTION]: The core workflow in SKILL.md instructs the agent to validate documentation by executing shell commands including python -m doctest, pytest --doctest-modules, and tsc --noEmit. These commands execute code blocks found within the documentation or project files to ensure accuracy.
[EXTERNAL_DOWNLOADS]: The skill recommends installing and using several standard technical documentation and linting tools via official package registries. This includes Python packages like pydocstyle, interrogate, and linkchecker, as well as Node.js tools like @redocly/cli, eslint-plugin-jsdoc, and static site generators like Docusaurus and MkDocs (references/coverage-reports.md, references/documentation-systems.md).
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it is designed to ingest and execute code snippets located within docstrings or documentation files.
Ingestion points: Source code files and documentation markdown provided by the user for processing (SKILL.md).
Boundary markers: There are no explicit instructions or delimiters used to separate the documentation content from the execution context to prevent malicious snippets from being executed.
Capability inventory: The agent has the capability to run local shell commands (python, pytest, npx) as part of its validation workflow (SKILL.md).
Sanitization: No sanitization, validation, or sandboxing of the code extracted from documentation is performed before the execution step.

code-documenter