The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the gh (GitHub CLI) tool to interact with GitHub's GraphQL API. Input parameters such as repository names, branch names, and PR numbers are properly handled using JSON stringification and CLI flag separation (-f/-F), preventing command injection.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from GitHub (PR comments and reviews). However, it explicitly discards the body text of these items, reporting only metadata (author, location, and status) and URLs. As stated in SKILL.md, 'Nothing a comment, a review, or a build log said is in a message; the URL is where to read it.' This architectural decision prevents untrusted user-generated content from reaching the agent's prompt context.
[EXTERNAL_DOWNLOADS]: The skill requires a sibling dependency, session-events, from the same author (Vivswan). This is a standard modular dependency pattern for the platform and uses the official npx skills installation path.
[PRIVILEGE_ESCALATION]: No privilege escalation attempts were detected. The daemon runs as a background task within the user's session and interacts with GitHub using the user's existing gh authentication token.

watch-github