The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill's setup process involves installing Node.js dependencies via a standard registry and mentions an install.sh script for automated configuration of the MCP server environment.
[COMMAND_EXECUTION]: The documentation provides instructions for the manual compilation and execution of the MCP server using npm install and npm run build, which are routine developer operations for this type of integration.
[PROMPT_INJECTION]: The skill has an inherent exposure to indirect prompt injection (Category 8) due to its core function of processing external, untrusted content from GitLab merge requests.
Ingestion points: Untrusted data from GitLab (titles, descriptions, and code diffs) is brought into the agent's context through the get_merge_request and get_merge_request_diff tools in mcp-server/src/index.ts.
Boundary markers: Absent. The skill instructions in SKILL.md do not currently define explicit delimiters to isolate external content from the agent's internal analysis logic.
Capability inventory: The agent uses the fetched data to perform stylistic, logic, security, and performance analyses and summarizes its findings in a report.
Sanitization: The skill does not perform pre-processing or filtering of the fetched content, relying instead on the agent's internal reasoning during the analysis phase.

code-review