The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the gh (GitHub) CLI, git, and npm to automate repository management.
It executes gh repo list and gh api to discover and inspect organization resources.
It performs local file operations in /tmp/ to clone and modify repository contents.
It automates the creation of pull requests and issues using gh pr create and gh issue create.
[EXTERNAL_DOWNLOADS]: The skill clones external code from GitHub repositories using the gh repo clone command.
This behavior is central to the skill's purpose of cross-repository coordination.
It utilizes standard GitHub APIs and official CLI tools for all network operations.
[REMOTE_CODE_EXECUTION]: The skill executes npm install, npm test, and npm audit fix on code cloned from external repositories.
This constitutes execution of potentially untrusted code found in the target repositories.
The documentation includes examples of running npx ruv-swarm, which is a package associated with the vendor's ecosystem.
[PROMPT_INJECTION]: The skill includes a high-risk surface for indirect prompt injection (Category 8).
Ingestion points: Reads package.json and CLAUDE.md files from GitHub repositories, and processes output from gh search.
Boundary markers: No explicit delimiters are used to separate untrusted repository content from internal agent instructions.
Capability inventory: Full access to shell execution (bash), package management (npm), and repository modification (git push, gh api PUT).
Sanitization: Uses jq to parse structured JSON data, reducing the risk of shell injection during metadata processing.

qe-github-multi-repo