stitch-mcp-delete-project
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the stitch-mcp tool suite to perform project deletion. While destructive, this behavior is the primary intended function and is restricted by the allowed-tools configuration.
- [INDIRECT_PROMPT_INJECTION]: The skill handles user-provided project identifiers. The potential risk of malicious data ingestion is mitigated by a mandatory human-in-the-loop confirmation process. 1. Ingestion points: Project identifiers provided in user requests (SKILL.md). 2. Boundary markers: Instructions explicitly require the agent to ask for user confirmation before tool invocation. 3. Capability inventory: delete_project, get_project, list_projects. 4. Sanitization: Instructions enforce strict projects/ID path formatting.
Audit Metadata