The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data in the form of git patches and diffs from GitLab Merge Requests. Malicious instructions embedded within the code changes or comments could potentially influence the LLM to generate a deceptive or harmful summary.
Ingestion points: Output from the glab mr diff command or the MCP patch retrieval tool (referenced in SKILL.md).
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the diff content as untrusted data or to ignore embedded instructions within that content.
Capability inventory: The skill uses shell command execution via the glab CLI and MCP tools to write data (updating MR descriptions) to a remote repository.
Sanitization: No sanitization or filtering of the diff content is performed before processing.
Mitigation: The skill includes a mandatory human-confirmation step ("Always show the proposed description and get user confirmation before writing to the remote"), which serves as a critical defense against automated exploitation.
[COMMAND_EXECUTION]: The skill uses the glab CLI tool to perform repository operations. While these are standard development tasks, executing commands derived from processed summaries (like the --description flag) requires the user to carefully review the output to prevent unintended command or argument injection.

update-mr-desc