The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates reading external data from GitLab, such as issue descriptions and merge request titles, which provides an attack surface for indirect prompt injection. Ingestion points: Tools such as GITLAB_LIST_PROJECT_ISSUES, GITLAB_GET_PROJECT_MERGE_REQUESTS, and GITLAB_LIST_REPOSITORY_COMMITS (referenced in SKILL.md). Boundary markers: There are no instructions to the agent to isolate or treat data from these tools as potentially untrusted. Capability inventory: The skill has extensive write and update capabilities, including GITLAB_CREATE_PROJECT_ISSUE, GITLAB_UPDATE_PROJECT_ISSUE, and GITLAB_CREATE_REPOSITORY_BRANCH (referenced in SKILL.md). Sanitization: No specific data validation or sanitization requirements are defined for inputs fetched from GitLab.
[EXTERNAL_DOWNLOADS]: The skill setup requires the user to configure an external MCP server endpoint at https://rube.app/mcp. Evidence: The documentation in SKILL.md directs users to add the service as an MCP server in their client configuration.

gitlab-automation