The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local shell commands including git for repository management and glab (GitLab CLI) for API interactions such as viewing merge requests, fetching diffs, posting comments, and merging.\n- [DATA_EXFILTRATION]: No malicious data exfiltration patterns were detected. The skill interacts only with the user's local repository and the configured GitLab instance via official CLI tools.\n- [PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from code diffs and merge request descriptions, which presents a surface for indirect prompt injection.\n
Ingestion points: Merge request descriptions (glab mr view) and cumulative code diffs (git diff, glab mr diff).\n
Boundary markers: The instructions include explicit directives to verify findings against the actual source code and strictly forbid the agent from automatically committing or pushing fixes.\n
Capability inventory: The agent can modify local files (auto-fix) and perform GitLab operations (approve, merge, comment) using the glab utility.\n
Sanitization: A mandatory verification phase (Phase 2.5) is implemented to ensure that identified issues are valid and reachable in the source code before any action is taken.\n- [SAFE]: The skill demonstrates security best practices for automated development tools by enforcing human-in-the-loop verification and checking for pre-conditions like CI pass and approvals before performing sensitive actions like merging.

gitlab-review