{"uid":"cap_cbHMQqMviKe3ij8-1e7B8","slug":"agent-workflow-red-team-analyzer-9ec40094","name":"Agent Workflow Red Team Analyzer","description":"A public journal about ideas, systems, experiments, projects, and learning.","url":"https://api.iamjuliand.com/v1/agent-workflow/red-team","method":"POST","headers":{},"bodySchema":null,"responseSchema":{"type":"json","example":{"model":"minimax/minimax-m3","usage":{"costUsd":0.00737,"promptTokens":1850,"completionTokens":920},"attacks":[],"summary":"The workflow can confuse untrusted ticket content with instructions.","findings":[],"riskScore":72,"mitigations":[],"dataHandling":{"toolsExecuted":false,"secretsRedacted":0,"externalProvider":"OpenRouter"},"deterministic":false,"productVersion":"1.0.0","staticAnalysis":{"policyGaps":["No explicit rule separates untrusted content from instructions."],"approvalGaps":[],"highImpactTools":["send_email"]},"regressionTests":[]}},"example":null,"exampleRequest":null,"tags":["x402"],"displayCostAmount":"0.12","displayCostAsset":"USDC","priceDynamic":false,"priceHint":null,"priceStatus":"priced","priceSource":"registry","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":"$0.12/call","primary":{"kind":"static","protocol":"x402","network":"base","amountUsd":"0.12","per":"call","confidence":"exact"},"accepted":[{"kind":"static","protocol":"x402","network":"base","amountUsd":"0.12","per":"call","confidence":"exact"}]},"paymentMethods":[{"uid":"pm_qf6Z-bHC1pJY3Ih0ktr7w","protocol":"x402","methodType":"crypto","chain":"base","mode":"charge","costAmount":"0.12","costPer":"request","priority":0,"asset":null,"unit":"request","depositMicros":null,"planRef":null}],"brandName":null,"brandSlug":null,"brandBaseUrl":null,"brandDocsUrl":null,"whatItDoes":"Performs automated red-team security analysis of an AI agent workflow, identifying prompt injection risks, policy gaps, and high-impact tool vulnerabilities.","exampleAgentPrompt":"Can you red-team my agent workflow for security vulnerabilities? It uses tools like send_email and processes untrusted ticket content — I want to know if it's vulnerable to prompt injection, what policy gaps exist, and what the overall risk score is.","exampleUseCases":[{"title":"Audit customer support agent pipeline","prompt":"Red-team my customer support agent — it reads user-submitted tickets and can call send_email and update_ticket tools. Tell me if untrusted ticket content could be mistaken for instructions, flag any policy gaps, and give me a risk score."},{"title":"Pre-deployment security check for coding agent","prompt":"Before I deploy my coding agent that reads GitHub issues and can run shell commands, can you analyze it for prompt injection risks and tell me which tools are high-impact and what mitigations I should put in place?"},{"title":"Ongoing regression testing for AI workflow","prompt":"I've updated my agent workflow and want to make sure I haven't introduced new vulnerabilities — run a red-team analysis and generate regression tests so I can catch prompt injection issues in CI."}],"resultDescription":"Returns a JSON object containing: a risk score (0-100), a natural language summary of vulnerabilities found, arrays of discovered attacks and findings, static analysis results (policy gaps, approval gaps, high-impact tools), suggested mitigations, regression test cases, and metadata about the model used, token usage, cost, and whether external tools were executed or secrets were redacted.","failureModes":["Workflow definition is malformed or missing required fields — returns error with missing field details","Ambiguous or underspecified policy rules lead to incomplete static analysis — findings array may be empty or misleading","Non-deterministic LLM behavior means repeated calls may yield different risk scores or findings","High complexity workflows may hit token limits, truncating analysis","External provider (OpenRouter) unavailability causes request failure"],"whenToPreferThis":"Choose this endpoint when you need automated, LLM-powered red-team analysis of an AI agent workflow specifically for prompt injection and instruction-confusion vulnerabilities. It is purpose-built for agentic pipelines with tool use and untrusted content ingestion, returning structured risk scores and actionable mitigations — prefer it over generic security scanners or manual review when you want a fast, quantitative security assessment of an AI agent.","instructions":null,"reviewSummary":null,"reviewSummaryHighlights":null,"reviewSummaryConcerns":null,"reviewSummaryGeneratedAt":null,"activationCount":0,"lastUsedAt":null,"lastSuccessfullyRanAt":null,"lastHealthCheckAt":"2026-09-14T00:32:36.159Z","isFirstParty":false}