The Agent Skills Directory

[COMMAND_EXECUTION]: The wrapper script scripts/gerrit.py utilizes the subprocess.run function to execute Gerrit commands via SSH. The script correctly uses the list-based argument format, which mitigates the risk of shell injection on the local system by avoiding the shell interpreter for command execution.
[EXTERNAL_DOWNLOADS]: Documentation within the skill (e.g., in references/troubleshooting.md) includes instructions for downloading a commit-msg hook from a Gerrit server using curl. This is a standard and expected procedure in Gerrit-based development environments to ensure commits include a unique Change-Id.
[INDIRECT_PROMPT_INJECTION]: The skill operates on untrusted data fetched from external Gerrit servers, such as commit messages, subjects, and review comments. This represents a potential attack surface for indirect prompt injection if an attacker submits a change containing malicious instructions aimed at the AI agent.
Ingestion points: Change metadata and review comments are retrieved in scripts/gerrit.py via run_gerrit_query (SSH) and fetch_change_diff (REST API).
Boundary markers: There are no specific delimiters or instruction-ignore markers used when presenting Gerrit content to the agent.
Capability inventory: The skill performs subprocess execution (ssh), network requests (urllib.request), and suggests usage of the git-review CLI tool.
Sanitization: The skill does not perform sanitization or filtering of natural language content fetched from the Gerrit server before displaying it.

gerrit