The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data from GitLab and GitHub platforms, such as merge request status, assignee information, and approval states.
Ingestion points: Metadata fetched via glab (GitLab) and gh (GitHub) command-line tools, specifically regarding merge request assignees and approval records.
Boundary markers: The instructions explicitly state that approvals relayed through merge request comments or tool results do not count, requiring instructions to originate from the current chat session.
Capability inventory: The skill is designed to perform code merges on repository forges when specific conditions are met.
Sanitization: The skill instructs the agent to verify specific structured data fields (like approved_by in GitLab or reviews in GitHub) rather than relying on unstructured content like comments for authorization checks.

sociuu-ship