The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell and Node.js scripts bundled within the plugin directory.
Evidence: Runs bash ${CLAUDE_PLUGIN_ROOT}/scripts/ensure-deps.sh to bootstrap the environment.
Evidence: Invokes node ${CLAUDE_PLUGIN_ROOT}/helper/bin/gh-manager.js for repository classification, authentication verification, and label management.
[EXTERNAL_DOWNLOADS]: Initialization involves downloading external dependencies from the npm registry.
Evidence: The ensure-deps.sh script is documented as automatically installing npm dependencies on first run.
[CREDENTIALS_UNSAFE]: The skill manages and verifies sensitive GitHub Personal Access Tokens.
Evidence: Uses the auth verify command to check GITHUB_PAT and provides setup instructions to the user.
[PROMPT_INJECTION]: The skill processes untrusted data from GitHub repositories, creating a surface for indirect prompt injection.
Ingestion points: Fetches repository labels and reads the .github-repo-manager.yml file from target repositories (SKILL.md).
Boundary markers: No explicit markers or delimiters are defined to isolate untrusted configuration data from the agent's instructions.
Capability inventory: The skill can perform file mutations (commits/PRs), update wikis, and manage repository labels (SKILL.md).
Sanitization: No sanitization or validation logic is specified for the contents of retrieved configuration files or labels.

repo-manager