$ loading_
Run safety-invariant checks on structured outputs to protect locked fields.
Copy the install command and let the AI configure it · recommended for beginners
No copy-paste install info for "invarianteval-mcp" yet — see the docs or source repo.
Use check_invariants to validate this structured extraction result, confirm that no locked fields were auto-filled by the model, and return any violations.
A deterministic validation result showing whether any fields violated the invariants.
Call list_invariants and list the available invariant checks so I can choose which rules to apply.
A list of available invariants to help the user understand supported safety checks.
Before submitting the result, use check_invariants to inspect all locked fields that require human confirmation and ensure the model did not fill them in.
An audit result indicating which fields are safe to keep and which require human handling.
Developers or researchers can use it to deterministically validate AI-generated structured outputs after extraction. It is especially useful when models must not silently complete critical fields.
When certain fields must only be filled after human confirmation, this tool can detect whether the model populated locked fields anyway. This helps block risky outputs before submission or storage.
Product managers or engineering teams can integrate it into AI workflows to apply consistent invariant checks to structured outputs. It fits well as a safety audit step for high-risk tasks.
It performs safety-invariant checks on structured outputs. Its main purpose is to ensure locked fields are never auto-filled by the model without human confirmation.
The description says it provides two tools: check_invariants and list_invariants. The first runs validations, and the second lists available invariants.
The provided material does not include installation steps, runtime requirements, or key requirements. See the source repository for details.
Check financial statements for footing and tie-out consistency issues quickly.
Validate AI agent behavior contracts and detect regressions through baseline comparisons.
Validate AI agent outputs for hallucinations, scope compliance, and quality scoring.
Evaluate AI safety classifier robustness against decomposition, obfuscation, and multi-agent attacks.
Validate JSON and SQL schemas with AI-driven customizable lint rules.
Analyze contract safety risks, flag dangerous clauses, and suggest improvements.