The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from GitHub, GitLab, and Bitbucket issue bodies and comments via CLI tools. This external content is subsequently used to influence the behavior of multiple sub-agents during the root cause analysis and fix design phases.
Ingestion points: SKILL.md (Phase 1) and references/fix-phases.md (Phase 1) use gh issue view to read issue descriptions and comments.
Boundary markers: The skill references a rule untrusted-input-quarantine.md and explicitly labels the issue body as "untrusted input".
Capability inventory: The skill has broad capabilities including Bash, Write, Edit, and the ability to spawn multiple background agents via the Agent tool.
Sanitization: The instructions suggest using a "read-only reader" to extract structured facts from the raw issue body before processing.
[COMMAND_EXECUTION]: Extensive reliance on the Bash tool for repository management, git operations (worktrees, branches, commits), and running the gh (GitHub), glab (GitLab), and bb (Bitbucket) command-line interfaces.
[EXTERNAL_DOWNLOADS]: The skill references the installation of the portless tool via npm and identifies a dependency on the yg-mcp-core package, which is hosted on a vendor-specific private repository at pypi.yonyon.ai.

fix-issue