The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest resource names provided by the user and pass them to a CLI search tool to resolve IDs, creating a surface for potential injection attacks.
[INGESTION_POINTS]: User-supplied resource names (SKILL.md).
[BOUNDARY_MARKERS]: The search command wraps the input in double quotes in the examples, but there is no explicit instruction to sanitize or handle special characters that could break out of the quote.
[CAPABILITY_INVENTORY]: The skill utilizes the celigo CLI tool for resource searching (SKILL.md).
[SANITIZATION]: There are no instructions for validating, escaping, or filtering the user-provided search terms before interpolation into the command line.
[SAFE]: The skill is otherwise focused on security best practices, including detailed sections on least-privilege scoping, secret management (never pasting tokens into chat), and proper token rotation and revocation lifecycles.
[SAFE]: No malicious obfuscation, persistence mechanisms, or unauthorized privilege escalation patterns were detected in the instructions.

managing-api-tokens