The Agent Skills Directory

[COMMAND_EXECUTION]: The generator script (assets/gen-docs.mjs) uses execSync to run git ls-files -z. This command is hardcoded and used solely to identify files tracked by Git, ensuring consistent documentation generation across different development environments.
[INDIRECT_PROMPT_INJECTION]: The skill processes repository source code to generate documentation summaries, which presents an attack surface where maliciously crafted content (such as SQL comments or file paths) could be reflected in the output files.
Ingestion points: The script assets/gen-docs.mjs reads the content of repository files via readFileSync to build inventories and metrics.
Boundary markers: The generated Markdown files include a banner indicating they are automated, but do not employ specific delimiters to isolate potentially malicious user-authored content within the code.
Capability inventory: The skill has permissions to read files, write to the docs/reference/ directory, and execute the git CLI.
Sanitization: The script performs basic extraction and truncation for SQL comments but does not escape content before writing to Markdown, meaning an attacker could theoretically inject Markdown formatting to mislead users or agents.
[EXTERNAL_DOWNLOADS]: The GitHub Action template (assets/docs-generate.yml) references official, well-known actions (actions/checkout and actions/setup-node) to prepare the runner environment. These are standard components for secure CI/CD workflows.

autowiki