The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from GitLab (e.g., merge requests, issue comments, and labels) which could contain malicious instructions intended to influence the agent's behavior.
Ingestion points: Tools such as read_merge_request, list_project_labels, and list_milestones (documented in references/tool-catalog.md) ingest content from external GitLab repositories.
Boundary markers: The skill implements mandatory confirmation gates (requirements_confirmed, projects_confirmed, issues_creation_confirmed) described in SKILL.md that require explicit human approval before the agent takes significant actions.
Capability inventory: The skill has access to tools capable of modifying the environment, including create_issue, set_issue_labels, add_issue_comment, and delete_issue (documented in references/tool-catalog.md).
Sanitization: The instructions mandate the use of generate_issue_payload for non-RF execution tasks rather than allowing the agent to hand-craft payloads, providing a layer of structured data validation.
[COMMAND_EXECUTION]: The skill directs the agent to execute specific shell commands to extract environment metadata for project discovery.
Evidence: The instructions in SKILL.md require running git remote get-url origin to identify the project path before sending it to the MCP server.

mcp-gitlab-usage