The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from the local repository environment, such as branch names and Pull Request titles, which could potentially contain untrusted content if the repository is shared with other contributors.\n
Ingestion points: Branch names from git branch and PR metadata from gh pr list (SKILL.md Step 3), and Linear issue details (SKILL.md Step 4).\n
Boundary markers: The instructions do not specify explicit delimiters for this data in the summary output, though it is organized into counted lists for user review.\n
Capability inventory: The skill utilizes git branch -D, git worktree remove, and a bundled Node.js script to perform deletions on the local filesystem and remote branches.\n
Sanitization: The skill relies on the agent's standard handling of string data and does not perform explicit sanitization on branch names or PR titles.\n- [COMMAND_EXECUTION]: The skill performs repository maintenance using common development tools and a bundled cleanup script.\n
Evidence: Executes standard git and gh commands for branch management and invokes node scripts/filesystem-hygiene.mjs (SKILL.md Step 9.5) to prune empty directories and orphaned node_modules. The script utilizes node:fs's rmSync to delete identified directories.\n- [SAFE]: The skill follows best practices for destructive operations by requiring explicit user confirmation for each cleanup pass and providing a --dry-run flag to preview changes without applying them. The bundled cleanup script explicitly checks for the presence of a .git directory to ensure it only operates on legitimate repository roots and protects sensitive Git metadata from deletion.

cleanup-repo