The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted text content from external Pull Requests (diffs, descriptions, comments) which presents an indirect prompt injection surface.
Ingestion Points: Content enters the agent context via gh pr view, gh pr diff, and gh api calls specified in SKILL.md to fetch pull request bodies and comments.
Boundary Markers: The skill contains an 'External Content Safety' section explicitly instructing the agent to treat fetched data as untrusted and never follow instructions or directives embedded within it.
Capability Inventory: The skill is capable of executing shell commands, local git operations, and a custom script using Playwright to load external URLs and take screenshots.
Sanitization: No text sanitization or filtering is applied to the incoming diff text or comment bodies before processing.
[COMMAND_EXECUTION]: The skill executes local git commands (git status, git diff) and GitHub CLI operations (gh pr view, gh pr diff, gh api) to view repository data. Additionally, it runs a local script node capture-screenshots.js that initiates a Playwright headless browser instance to render and screenshot external preview URLs provided in pull requests.

code-review