The Agent Skills Directory

[SAFE]: No malicious patterns or security risks were detected in the skill instructions or workflow.
[SAFE]: Data Ingestion Security: The skill acknowledges that data from external support tickets is untrusted and provides explicit instructions to ignore any embedded commands or prompts.
[SAFE]: Human-in-the-loop Verification: The workflow requires a mandatory user approval step ("Present for review") before the agent can execute the MCP tool to post a note, preventing autonomous data exfiltration or unintended message posting.
[SAFE]: Input Validation: The skill includes strict regex validation (/^[0-9a-f]{24}$/i) for the ticket identifier to prevent injection or malformed data usage.
[SAFE]: Least Privilege: The skill relies on an MCP server for authentication, avoiding the need for hardcoded API keys or environment variable access by the agent.

gleap-responder