The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data from GitHub (issues, pull requests, and CI logs) which could contain adversarial instructions, though this is a standard risk for development tools.
Ingestion points: The agent reads content from GitHub using gh issue view, gh pr view, and gh run view --log-failed (SKILL.md).
Boundary markers: The skill lacks explicit instructions to ignore embedded commands in external text, though it recommends summarizing and formatting results which provides some implicit context separation.
Capability inventory: The skill has the ability to create and merge pull requests, close issues, and run GitHub API queries via the gh CLI.
Sanitization: No specific sanitization or filtering logic is applied to the retrieved text before it enters the agent's context.
[COMMAND_EXECUTION]: The skill uses the official GitHub CLI (gh) to perform repository operations.
Evidence: Standard commands like gh issue create, gh pr merge, and gh search are utilized throughout SKILL.md.
Safety: The instructions explicitly mandate seeking user confirmation before performing destructive operations like merging PRs or closing issues, minimizing the risk of accidental or unauthorized changes.

github