The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill uses dynamic module loading to manage its internal script dependencies. Specifically, scripts/audit-doc-health.py and tests/test_doc_governance_scripts.py utilize importlib.util to define and execute code from sibling scripts (check-doc-links.py and trim-revision.py). While these paths are restricted to the script's own directory, dynamic loading from computed paths is a vector often used for code injection.
[COMMAND_EXECUTION]: The test suite (tests/test_doc_governance_scripts.py) uses the subprocess module to execute the scripts/scaffold-doc.sh shell script. This allows the skill to programmatically run shell commands to verify its template generation logic.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and audit user-provided Markdown documentation, creating an attack surface for indirect prompt injection.
Ingestion points: Files are read from the local file system using Path.read_text() in scripts/audit-doc-health.py, scripts/check-doc-links.py, scripts/generate-llms-txt.py, and scripts/trim-revision.py.
Boundary markers: The system does not explicitly define delimiters or 'ignore' instructions for the content it audits.
Capability inventory: The skill has the ability to write to the file system (modifying .md files in trim-revision.py and creating llms.txt in generate-llms-txt.py) and perform command execution via the test environment.
Sanitization: Content is processed by various regular expressions to extract metadata and links, but it lacks structured sanitization to prevent the agent from obeying embedded instructions found within parsed documentation content.

doc-governance