The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and process untrusted data from source code, configuration files, and comments to generate documentation. This creates an attack surface where malicious instructions embedded in the code (e.g., in docstrings or comments) could influence the documentation expert's output.
Ingestion points: SKILL.md instructions specify extracting information from 'code, configs, and comments'. resources/implementation-playbook.md provides Python examples using ast.parse to read local files.
Boundary markers: The skill lacks explicit instructions to use delimiters or 'ignore' directives when processing the content of these files, although it does include a safety warning to avoid exposing secrets.
Capability inventory: The skill possesses the capability to write files (documentation artifacts) and suggests configuring CI/CD pipelines (shell execution).
Sanitization: No explicit sanitization or filtering logic is described for the extracted content before it is interpolated into documentation templates.
[COMMAND_EXECUTION]: The resources/implementation-playbook.md file contains several examples of shell commands for installation (pip install), environment setup (git clone, python -m venv), and documentation building (sphinx-build, pytest). These are intended as templates for the user or agent to configure project environments.
[DYNAMIC_EXECUTION]: The provided Python scripts in the playbook use the ast (Abstract Syntax Tree) module to parse and analyze Python code dynamically. While ast.parse is safer than eval(), it is used to programmatically inspect and extract data from external code files.

code-documentation-doc-generate