The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from the Sentry UI which may contain untrusted strings, representing a theoretical injection surface.
Ingestion points: Captures screenshots of pages, drawers, and modals in the Sentry web application (SKILL.md).
Boundary markers: None present for the content captured within images.
Capability inventory: Executes gh CLI commands to edit pull requests and comments; writes to local files like pr-body.md (SKILL.md).
Sanitization: The skill mandates strict redaction of sensitive values (emails, slugs, IDs) and a manual audit for credentials (tokens, auth headers, cookies) before uploading (SKILL.md, Step 3).
[COMMAND_EXECUTION]: The skill utilizes the GitHub CLI (gh) to perform its primary functions.
Evidence: Instructions include using gh pr create, gh pr edit, and gh --attach to manage pull request metadata and upload image assets (SKILL.md, Step 5).
[DATA_EXFILTRATION]: The skill transmits captured image data to external servers.
Evidence: Images are uploaded to GitHub via the gh --attach command. This targets a well-known service and is a core component of the skill's intended workflow (SKILL.md, Step 5).

pr-screenshots