The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from GitLab merge requests, including diffs and comments, which could be used as a vector for indirect prompt injection. * Ingestion points: The skill executes glab mr view and git diff to retrieve external content in SKILL.md. * Capability inventory: The skill has access to the Bash tool for shell commands, the glab CLI for GitLab API interactions, and the Skill tool for invoking other agents. * Boundary markers: There are no explicit instructions or delimiters used to separate untrusted data from the agent's instructions. * Sanitization: No sanitization logic is specified for the data retrieved from GitLab.
[COMMAND_EXECUTION]: The skill uses the Bash tool to run git and glab commands. While necessary for functionality, this provides a capability that could be targeted if a successful prompt injection occurs.

review-mr