The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to generate a bash script (collect-ai-metrics.sh) from templates and execute it to extract project data via git and the GitHub (gh) CLI.
[EXTERNAL_DOWNLOADS]: The generated HTML dashboard template fetches font assets from Google's official font service.
[DATA_EXFILTRATION]: The skill aggregates potentially sensitive information, including absolute local file paths, git remote URLs, and developer email patterns from commit history, into Markdown, JSON, and HTML reports.
[PROMPT_INJECTION]: The skill processes untrusted local project files and git metadata to populate its reports, creating a surface for indirect prompt injection if those files contain malicious instructions.
Ingestion points: Local files such as CLAUDE.md, README.md, package.json, and git commit trailers (Co-Authored-By).
Boundary markers: No delimiters or protective warnings are used when the agent processes these inputs.
Capability inventory: The skill has access to shell execution, broad file system reading, and network access through the GitHub API.
Sanitization: Data is formatted using jq, but there is no explicit sanitization of the string content extracted from untrusted files.

project-inventory