The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/audit-doc-health.py uses importlib.util to dynamically load sibling scripts (such as check-doc-links.py and trim-revision.py) from the local file system. This is used to orchestrate the skill's internal auditing functions and does not ingest untrusted remote code.
[COMMAND_EXECUTION]: The script scripts/scaffold-doc.py uses subprocess.run to execute local helper scripts like manage-doc-index.py and manage-backlog.py to automate documentation indexing. Additionally, the test suite (tests/) uses subprocess.run to perform integration testing of the CLI tools. These executions are scoped to the skill's own directory and are necessary for its documented functionality.
[INDIRECT_PROMPT_INJECTION]: The skill represents an attack surface for indirect prompt injection because it processes untrusted Markdown content (reading files to check links, audit health, and generate summaries). However, the scripts operate locally and do not exfiltrate data or perform dangerous operations based on the content of the files being audited.
Ingestion points: audit-doc-health.py, check-doc-links.py, and generate-llms-txt.py read the content of Markdown files in the targeted documentation directory using Path.read_text().
Boundary markers: The skill encourages the use of managed markers like  to delimit generated content, but these are primarily for maintenance rather than security boundaries.
Capability inventory: The skill possesses capabilities for local file reading/writing and local subprocess execution (Python scripts only).
Sanitization: The scripts use shlex.quote to sanitize project names before suggesting shell commands to the user and slugify to sanitize filenames, reducing the risk of local command injection.

doc-governance