{"uid":"cap__t2L_sOSf7cxCnYtJocyv","slug":"verdict-tool-output-guard-2127f8e7","name":"Verdict Tool Output Guard","description":"Tool results (MCP servers, APIs, search, file reads) are untrusted content the agent re-reads. Detects a result that gives orders to the agent, strays from what the tool should return or asks to send data (tool poisoning). Returns allow, review or block.","url":"https://verdict.kechlab.com/v1/tool-output-guard?utm_source=zero.xyz","method":"POST","headers":{},"bodySchema":{"type":"object","properties":{"output":{"type":"string","description":"the result returned by the tool"},"purpose":{"type":"string","description":"(optional) what the tool is supposed to return"},"tool_name":{"type":"string","description":"(optional) tool name"}}},"responseSchema":null,"example":null,"exampleRequest":null,"tags":["x402"],"displayCostAmount":"0.0133","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":"$0.0133/call","primary":{"kind":"static","protocol":"x402","network":"base","amountUsd":"0.0133","per":"call","confidence":"exact"},"accepted":[{"kind":"static","protocol":"x402","network":"base","amountUsd":"0.0133","per":"call","confidence":"exact"}]},"paymentMethods":[{"uid":"pm_Ff54zud_xfVxikCrbuDZl","protocol":"x402","methodType":"crypto","chain":"base","mode":"charge","costAmount":"0.0133","costPer":"request","priority":0,"asset":"0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913","unit":"request","depositMicros":null,"planRef":null}],"brandName":null,"brandSlug":null,"brandBaseUrl":null,"brandDocsUrl":null,"whatItDoes":"Detects prompt injection and tool-poisoning attacks in tool/API results returned to an AI agent, returning allow, review, or block.","exampleAgentPrompt":"Before you process the result from that web-search tool, run it through Verdict's tool-output guard — the tool is called 'web_search' and it's supposed to return search snippets — check whether the output is safe or if it's trying to hijack you.","exampleUseCases":[{"title":"MCP server result safety check","prompt":"I just got a response back from an MCP file-reader tool called 'read_file' that's supposed to return file contents. Can you run it through the tool-output guard to make sure it doesn't contain any hidden instructions trying to take over your behavior before you use it?"},{"title":"Search result injection screening","prompt":"Before you summarize those search results from the Bing tool, pass each result through the tool-output guard — the tool is 'bing_search' and it should only return page snippets — so we can block any that are injecting commands."},{"title":"Third-party API response validation","prompt":"The weather API just returned this JSON blob — run it through Verdict's tool-output guard with the tool name 'weather_api' and the purpose 'return current temperature and forecast', and tell me if it's safe to act on or if something looks off."}],"resultDescription":"Returns a verdict of 'allow', 'review', or 'block' indicating whether the tool output is safe, suspicious, or contains prompt injection / tool-poisoning content. Also indicates whether the output strays from what the tool is supposed to return or attempts to exfiltrate data.","failureModes":["Ambiguous tool output may return 'review' requiring human decision","False positives on legitimate tool outputs that use imperative language","Missing 'purpose' field may reduce detection accuracy for context-dependent injections","Very long tool outputs may be truncated or cause latency spikes","Novel injection techniques not yet in the model's training may be missed"],"whenToPreferThis":"Use this endpoint when your AI agent consumes results from external tools, MCP servers, APIs, search engines, or file reads and you need to screen those results for prompt injection or tool-poisoning before the agent acts on them. Prefer this over general content moderation when the threat model is specifically adversarial tool outputs trying to redirect agent behavior, exfiltrate data, or issue unauthorized commands — rather than offensive language or NSFW content.","instructions":null,"reviewSummary":null,"reviewSummaryHighlights":null,"reviewSummaryConcerns":null,"reviewSummaryGeneratedAt":null,"activationCount":0,"lastUsedAt":null,"lastSuccessfullyRanAt":null,"lastHealthCheckAt":"2026-10-21T07:21:01.353Z","isFirstParty":false,"canonicalSlug":"verdict-tool-output-guard-2127f8e7"}