model-a-protocol
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to run the
tla-tsmodel checker CLI viapnpm. This is a legitimate use of project-local tooling for formal verification of protocol specifications within the@rat-stack/coreworkspace. - [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow where the agent processes and generates TLA+ specifications and configuration files. This represents a potential surface where instructions embedded in specs or documentation could influence agent behavior, though the risk is minimal in the intended developer-centric context.
- Ingestion points: The agent reads TLA+ specifications (
.tla), configurations (.cfg), and associated documentation (README.md) from the project'spackages/core/spec/directory. - Boundary markers: The skill does not define specific delimiters or instructions to disregard potential commands embedded within the formal specifications or code comments.
- Capability inventory: The agent is expected to execute the
tla-tsCLI tool and perform file operations (read/write) on the codebase to create and verify models. - Sanitization: No explicit sanitization or validation steps are described for the content of the TLA+ files before they are processed by the agent or the model checker tool.
Audit Metadata