The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple sources (GitHub links, external URLs, local file paths, and pasted text) in the skill-extractor sub-agent. This content is then used to drive downstream actions such as classification and GitHub issue creation.
Ingestion points: agents/extractor.md reads from user-provided URLs and file paths.
Boundary markers: The main SKILL.md requires the agent to display the extracted information to the user and wait for explicit confirmation/correction before proceeding to further steps.
Capability inventory: The skill uses gh issue create, gh project item-add, and gh api to write data to GitHub repositories.
Sanitization: Reliance is placed on human review of the extracted text; no automated sanitization is specified.
[COMMAND_EXECUTION]: The skill frequently executes shell commands via the GitHub CLI (gh) to manage issues and projects. Examples include gh issue create, gh project item-list, and gh project item-edit in agents/proposer.md and agents/deduplicator.md. It also uses gh api to push updates to repository files. These operations target the author's repositories.
[EXTERNAL_DOWNLOADS]: The skill-extractor sub-agent is designed to fetch content from remote URLs and GitHub repositories to parse skill descriptions. These targets are determined by user input at runtime.

skill-sourcer