The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local scripts using pnpm and bun (e.g., pnpm issues:project, bun mirror-prs.ts) and interacts with the system through the gh (GitHub CLI) for repository operations.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external, untrusted data from GitHub issues, pull requests, and comments, projecting them into the local workspace as structured markdown files. This creates an attack surface where malicious instructions embedded in GitHub issue bodies could influence the agent's context.
Ingestion points: The src/issues/projector.ts script fetches data from tracked repositories under orgs/**.
Boundary markers: No explicit delimiters or safety instructions are defined to separate ingested data from agent instructions.
Capability inventory: The skill can write files to the workspace (issues/org/<owner>/<repo>/) and perform GitHub repository actions (branches, PRs, issues).
Sanitization: There is no documented sanitization or filtering of the fetched Markdown content before it is projected into the workspace.
[CREDENTIALS_UNSAFE]: The skill relies on a GITHUB_TOKEN environment variable for authentication. While no hardcoded secrets are present and the use of placeholders like <token> is standard, the skill documentation encourages patterns where sensitive tokens are passed via environment variables.

github-integration