thermos
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The maintenance documentation includes shell commands for git, pnpm, and npx to facilitate manual updates and validation of the skill set. These commands are documentation for developers and are not executed automatically by the agent.
- [EXTERNAL_DOWNLOADS]: The skill references the official Cursor plugin repository (github.com/cursor/plugins.git) for upstream updates and utilizes npx for installation from the vendor's repository. These sources are well-known or vendor-affiliated.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from repository diffs and file contents (SKILL.md). This surface could allow instructions embedded in code to influence the subagents. The skill mitigates this by providing specific rubrics, using isolated subagent contexts, and requiring that findings be deduplicated and verified against source evidence before reporting.
Audit Metadata