{"uid":"cap_xp6lb15WHMOcpr0JsuERr","slug":"salt19-agent-plan-verifier-2b605db2","name":"SALT19 Agent Plan Verifier","description":"SALT19 is an independent applied AI systems lab behind EvoMind governed cognition, the MCP-native Agent Utility Grid, the ARCS research community, AeroClear UAS flight intelligence, and practical software for real-world work.","url":"https://api.salt19.com/v1/agent-plan-verifier","method":"POST","headers":{},"bodySchema":{"type":"object","properties":{"goal":{"type":"string","maxLength":4000,"minLength":1},"plan":{"type":"array","items":{"type":"string","maxLength":4000,"minLength":1},"maxItems":64,"minItems":1},"constraints":{"type":"array","items":{"type":"string","maxLength":4000,"minLength":1},"maxItems":32}}},"responseSchema":{"type":"json","example":{"score":1,"repairs":[],"verdict":"GO","unsafe_steps":[],"ordering_errors":[],"constraint_violations":[],"missing_preconditions":[]}},"example":null,"exampleRequest":null,"tags":["x402"],"displayCostAmount":"0.72","displayCostAsset":"USDC","priceDynamic":false,"priceHint":null,"priceStatus":"priced","priceSource":"probe","requiresHandshake":false,"reviewCount":0,"rating":{"score":"0.00","successRate":"0.00","reviews":0,"stars":null,"state":"unrated"},"availabilityStatus":"unknown","priceObserved":null,"sessionDeposit":null,"pricing":{"kind":"static","summary":"$0.72/call","primary":{"kind":"static","protocol":"x402","network":"base","amountUsd":"0.72","per":"call","confidence":"exact"},"accepted":[{"kind":"static","protocol":"x402","network":"base","amountUsd":"0.72","per":"call","confidence":"exact"}]},"paymentMethods":[{"uid":"pm_v5lae85ccNDtDXM9dj-z_","protocol":"x402","methodType":"crypto","chain":"base","mode":"charge","costAmount":"0.72","costPer":"request","priority":0,"asset":"0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913","unit":"request","depositMicros":null,"planRef":null}],"brandName":null,"brandSlug":null,"brandBaseUrl":null,"brandDocsUrl":null,"whatItDoes":"Validates an AI agent's step-by-step plan against a goal and optional constraints, returning a verdict, score, and itemized issues such as unsafe steps, ordering errors, and constraint violations.","exampleAgentPrompt":"Before we run this, check my plan: the goal is 'deploy the new microservice to production', and the steps are ['run unit tests', 'build Docker image', 'push to registry', 'apply Kubernetes manifest', 'notify team'] — also enforce the constraints that staging must pass before production and no direct DB migrations during deploy.","exampleUseCases":[{"title":"Pre-execution safety check for autonomous agent","prompt":"I have an AI agent about to take a series of actions to migrate our database. The goal is 'migrate customer data to new schema with zero downtime'. Steps are: ['backup current DB', 'run migration script', 'verify row counts', 'switch connection strings', 'drop old tables']. Constraints: 'no data loss allowed' and 'rollback must be possible at each step'. Can you verify this plan is safe and correctly ordered before we kick it off?"},{"title":"Constraint compliance audit for a procurement workflow","prompt":"We've got a purchasing workflow plan and need to make sure it follows company policy. Goal: 'procure 50 laptops for the new team'. Steps: ['get manager approval', 'submit purchase request', 'select vendor', 'place order', 'receive and log assets']. Constraints: 'approval must precede any purchase', 'vendor must be on approved list', 'orders over $10k need CFO sign-off'. Please score this plan and flag any violations."},{"title":"Debugging a broken agent plan before re-deployment","prompt":"My agent's plan keeps failing in production and I think there's a step ordering issue. Goal is 'send weekly performance report to stakeholders'. Steps: ['generate PDF report', 'fetch latest metrics', 'email PDF to stakeholders', 'archive report to S3']. Can you run this through the plan verifier and tell me what's wrong — missing preconditions, wrong order, anything?"}],"resultDescription":"Returns a JSON object with a numeric score (0-1), a verdict string (e.g. 'GO' or 'NO-GO'), and arrays identifying unsafe steps, ordering errors, constraint violations, missing preconditions, and suggested repairs. An empty arrays and score of 1 indicate a fully valid plan.","failureModes":["Plan array is empty or contains empty strings — returns validation error","Goal field missing or blank — request rejected with schema error","Steps exceed 64-item limit — input rejected","Constraints exceed 32-item limit — input rejected","Individual step or constraint string exceeds 4000 characters — truncation or rejection","Payment not included or insufficient — HTTP 402 response","Ambiguous or contradictory constraints may produce low-confidence verdicts"],"whenToPreferThis":"Choose this endpoint when an AI agent needs a structured, scored pre-execution safety check on a multi-step plan before committing to action — especially when constraints must be enforced, step ordering matters, or unsafe actions must be flagged. It is purpose-built for agentic workflows (MCP-native, EvoMind ecosystem) and returns machine-readable repair suggestions, making it superior to generic LLM prompting for plan validation in governed or safety-critical agent pipelines.","instructions":null,"reviewSummary":null,"reviewSummaryHighlights":null,"reviewSummaryConcerns":null,"reviewSummaryGeneratedAt":null,"activationCount":0,"lastUsedAt":null,"lastSuccessfullyRanAt":null,"lastHealthCheckAt":"2026-09-15T12:52:30.144Z","isFirstParty":false}