The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to run various shell commands for monorepo maintenance, including building packages (yarn build), running validation suites (yarn test:ci), and executing local scripts for releases (.github/scripts/create-releases.sh). These operations are consistent with the intended maintenance role within the vendor's own repository structure.
[INDIRECT_PROMPT_INJECTION]: The skill defines workflows for processing external data via GitHub issues (bug reports and feature requests), which represents a potential ingestion surface for malicious instructions.
Ingestion points: GitHub issues created using templates like bug_report.md and feature_request.md in the .github/ISSUE_TEMPLATE/ directory.
Boundary markers: The instructions recommend using specific templates and suggest the Design team review visual regressions, although they do not explicitly specify prompt-level delimiters or 'ignore' instructions for the automated bug-fixing agent.
Capability inventory: The agent has the capability to execute build, test, and release commands (yarn, npx eslint) and run local scripts that interact with the repository and CI/CD pipelines.
Sanitization: Risk is mitigated through mandatory human review of PRs generated by the bug-fixing agent and automated CI validation checks before code is merged or released.

repo-maintenance