setup-pstack
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEPERSISTENCEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PERSISTENCE]: The skill writes a persistent rule file to
~/.cursor/rules/pstack-models.mdcwith thealwaysApply: truemetadata. This ensures that model role assignments are consistently applied to future agent sessions. This behavior is consistent with the primary purpose of a configuration skill for the Cursor environment. - [INDIRECT_PROMPT_INJECTION]: The skill constructs an instruction file (a Cursor rule) that influences the behavior of the agent by defining which models are used for specific roles. The potential for malicious content injection into these instructions is mitigated by Step 4, which validates that every model slug written must be from a verified, detected set of available models.
- [COMMAND_EXECUTION]: The skill facilitates the invocation of other skills, specifically
/create-verification-skilland/create-formal-verification-skill, based on explicit user confirmation during the setup process.
Audit Metadata