{
  "schema": "varexis-public-evidence-record/0.3",
  "generated_at_utc": "2026-10-30T15:17:58.235744+00:00",
  "record_id": "VXI-EXT-ICE09-OPENAI-OUTPUT-TRIPWIRE-EXCEPTION-SESSION-EVIDENCE-01-v1",
  "record_kind": "external_reference",
  "title": "Does an output guardrail tripwire leave the same session evidence as a guardrail exception",
  "system": "OpenAI",
  "evidence_family": "vendor_evidence_review",
  "index_status": "PUBLIC_SAFE_SOURCE_REVIEW",
  "evidence_state": "PRIMARY_DOCUMENTATION_REVIEW_ONLY",
  "canonical_question": "Does an output guardrail tripwire leave the same session evidence as a guardrail exception?",
  "near_questions": [
    "Does an output tripwire persist completed tool calls?",
    "How does a guardrail exception differ from a rejected final answer?",
    "Which error wins if guardrail and session persistence both fail?",
    "Can rejected terminal-tool payloads be sanitized from persisted session evidence?"
  ],
  "supported_claims": [
    {
      "claim": "The captured Python SDK guide distinguishes output-tripwire rejection from a guardrail exception. A tripwire asks the configured session to retain replay-valid completed tool context while excluding rejected final output; rejected terminal-tool payloads can be replaced with a placeholder, or the response suffix discarded when safe sanitization is unavailable. A guardrail exception leaves the verdict unknown and requests persistence of completed final-turn items; a session-write error takes precedence. Retained call metadata can still contain arguments.",
      "reliance": "DOCUMENTED_CAPABILITY_ONLY"
    }
  ],
  "not_demonstrated_claims": [
    "Do not infer absent tool work from withheld output, assume rejected payloads are retained, or treat a guardrail exception as a rejection verdict. Inspect the actual session write and retained fields.",
    "Buyer implementation, live runtime enforcement, settled payment, verified deletion or certification are not established.",
    "No expanded capabilities or licences for existing INDEX executables."
  ],
  "public_summary": "The captured Python SDK guide distinguishes output-tripwire rejection from a guardrail exception. A tripwire asks the configured session to retain replay-valid completed tool context while excluding rejected final output; rejected terminal-tool payloads can be replaced with a placeholder, or the response suffix discarded when safe sanitization is unavailable. A guardrail exception leaves the verdict unknown and requests persistence of completed final-turn items; a session-write error takes precedence. Retained call metadata can still contain arguments.",
  "related_record_ids": [],
  "source_review": {
    "source_type": "official_external_documentation",
    "source_title": "OpenAI primary documentation: openai guardrails",
    "source_url": "https://openai.github.io/openai-agents-python/guardrails/",
    "checked_on": "2026-10-30",
    "candidate_record_id": "VXI-EXT-ICE09-OPENAI-OUTPUT-TRIPWIRE-EXCEPTION-SESSION-EVIDENCE-01-v1",
    "candidate_record_sha256": "8383a05d5edf414b218c40cf54936f81829ecd1c519773e2a0b9ac5874c621a9",
    "method": "Primary documentation reviewed for the published bounded statement.",
    "deployment_tested_by_varexis": false,
    "revalidation": "The delivered snapshot preserves the review date shown. Recheck current primary documentation and deployment context before relying on present applicability; payment does not trigger fresh research."
  },
  "answer_policy": {
    "classification": "DOCUMENTED_WITH_DEPLOYMENT_BOUNDARY",
    "bounded_answer": "The captured Python SDK guide distinguishes output-tripwire rejection from a guardrail exception. A tripwire asks the configured session to retain replay-valid completed tool context while excluding rejected final output; rejected terminal-tool payloads can be replaced with a placeholder, or the response suffix discarded when safe sanitization is unavailable. A guardrail exception leaves the verdict unknown and requests persistence of completed final-turn items; a session-write error takes precedence. Retained call metadata can still contain arguments.",
    "required_buyer_context": "SDK version; tripwire or exception; completed tool items; session write outcome. Buyer deployment not inspected.",
    "missing_evidence": "A broader claim needs evidence matching the intended use. Payment does not upgrade the evidence state."
  },
  "private_boundary": "Raw logs, thresholds, private paths, source code, protected mechanisms and reconstruction-level details are excluded from public and paid delivery.",
  "decision_profile": {
    "schema": "varexis-vendor-evidence-review/1",
    "review_status": "documentation_review_only",
    "applicability_status": "Rolling OpenAI Agents Python SDK guardrail documentation captured 2026-10-30; configured session, installed version, sanitization and actual persistence unverified.",
    "checked_on": "2026-10-30",
    "claim_to_check": "Does an output guardrail tripwire leave the same session evidence as a guardrail exception?",
    "documented_statement": {
      "status": "documented",
      "statement": "The captured Python SDK guide distinguishes output-tripwire rejection from a guardrail exception. A tripwire asks the configured session to retain replay-valid completed tool context while excluding rejected final output; rejected terminal-tool payloads can be replaced with a placeholder, or the response suffix discarded when safe sanitization is unavailable. A guardrail exception leaves the verdict unknown and requests persistence of completed final-turn items; a session-write error takes precedence. Retained call metadata can still contain arguments.",
      "source_ids": [
        "openai-guardrails"
      ]
    },
    "review_inference": {
      "status": "review_inference",
      "statement": "Does an output guardrail tripwire leave the same session evidence as a guardrail exception? Do not infer absent tool work from withheld output, assume rejected payloads are retained, or treat a guardrail exception as a rejection verdict. Inspect the actual session write and retained fields. Guardrail outcome changes retained session evidence beyond control-placement coverage. This is an operational review inference, not an observed deployment outcome.",
      "source_ids": [
        "openai-guardrails"
      ]
    },
    "buyer_evidence_required": [
      "SDK version",
      "tripwire or exception",
      "completed tool items",
      "session write outcome"
    ],
    "review_condition": {
      "status": "varexis_review_policy",
      "statement": "Do not infer absent tool work from withheld output, assume rejected payloads are retained, or treat a guardrail exception as a rejection verdict. Inspect the actual session write and retained fields."
    },
    "unknowns": "The buyer implementation, effective authority, source-data completeness and actual business outcomes were not inspected.",
    "unknown_finding": {
      "status": "unknown",
      "statement": "No buyer deployment or transaction was observed."
    },
    "evidence_sources": [
      {
        "source_id": "openai-guardrails",
        "url": "https://openai.github.io/openai-agents-python/guardrails/",
        "final_url": "https://openai.github.io/openai-agents-python/guardrails/",
        "captured_at_utc": "2026-10-30T14:58:02.844923+00:00",
        "status": 200,
        "content_type": "text/html; charset=utf-8",
        "bytes": 119635,
        "sha256": "644f53b4758f133dcc8b652c4dd489b8fbcb6c485de25486ec25885ebfce601d",
        "title": "OpenAI primary documentation: openai guardrails",
        "publisher": "OpenAI",
        "checked_on": "2026-10-30",
        "capture_status": 200,
        "claim_review_completed": true,
        "hash_basis": "SHA-256 of captured HTTP response-body bytes; not a signature or factual guarantee.",
        "version_context": "Rolling first-party documentation captured 2026-10-30; verify installed SDK/service and applicable account/version; deployment behaviour untested.; Rolling OpenAI Agents Python SDK guardrail documentation captured 2026-10-30; configured session, installed version, sanitization and actual persistence unverified.",
        "rights_disposition": "Original bounded paraphrase with source attribution. Captured source body remains private; no source-body redistribution. Vendor documentation rights, code licences and marks are not transferred.",
        "reviewed_sections": [
          "Parallel execution (default",
          "An output tripwire and an exception",
          "Terminal function-tool output needs additional handling"
        ]
      }
    ],
    "observed_deployment": false,
    "commercial_reliance_boundary": "Documentation supports only the stated boundary. Existing licensed checkers analyse supported supplied declarations locally; they do not inspect provider accounts, verify live behaviour, enforce decisions or certify compliance.",
    "paid_value": {
      "decision_context": "Does an output guardrail tripwire leave the same session evidence as a guardrail exception? Do not infer absent tool work from withheld output, assume rejected payloads are retained, or treat a guardrail exception as a rejection verdict. Inspect the actual session write and retained fields.",
      "free": "Complete source review, questions, provenance and limitations; free to read and save locally.",
      "single_request": "Unavailable for this publication-only record; historical saved retrieval is unchanged.",
      "evidence_receipt": "Unavailable for this publication-only record; no paid retrieval entitlement is added.",
      "excluded": [
        "Fresh source research",
        "Native provider integration or schema support",
        "Live inspection or runtime enforcement",
        "Payment settlement or verified deletion",
        "Certification or expanded licences"
      ],
      "integrity": "Hashes identify content versions; they do not prove factual truth or provide a signature."
    },
    "revalidation": {
      "trigger": "Recheck current primary documentation and the applicable API, service, account and configuration before reliance. No ongoing monitoring is included.",
      "specific_watch": "An output tripwire and an exception; Terminal function-tool output needs additional handling"
    },
    "buyer_relevance": "Does an output guardrail tripwire leave the same session evidence as a guardrail exception? Do not infer absent tool work from withheld output, assume rejected payloads are retained, or treat a guardrail exception as a rejection verdict. Inspect the actual session write and retained fields.",
    "refusal_conditions": [
      "Do not infer absent tool work from withheld output, assume rejected payloads are retained, or treat a guardrail exception as a rejection verdict. Inspect the actual session write and retained fields.",
      "Do not infer deployment conformance, authority or successful outcomes from source documentation."
    ],
    "decision_context": {
      "operator": "Agent integration or handoff reviewer",
      "scenario": "Does an output guardrail tripwire leave the same session evidence as a guardrail exception?",
      "consequence": "Do not infer absent tool work from withheld output, assume rejected payloads are retained, or treat a guardrail exception as a rejection verdict. Inspect the actual session write and retained fields.",
      "confidence": "High for the explicitly bounded captured clause; conflicts and version limits retained; deployment untested."
    },
    "novelty_review": {
      "candidate_id": "ICE09-222",
      "disposition": "EXTENSION",
      "nearest_record_id": "VXI-EXT-OAI-AGENTS-GR-03-v1",
      "nearest_record_url": "/varexis-index/records/agent-tool-guardrail-coverage.json",
      "nearest_record_sha256": "ab5bfc323f219f4d8765eba5b342dcb5f0263d9f70e6ccb7438cd6c463fe1881",
      "material_distinction": "Guardrail outcome changes retained session evidence beyond control-placement coverage.",
      "review_status": "PASS: source qualification plus independent reviewer challenge and resolved corrections"
    },
    "optional_selection": {
      "guide_id": "cross-agent-handoff-acceptance",
      "guide_url": "/index-guides/cross-agent-handoff-acceptance",
      "guide_manifest": "/varexis-index/decision-guides.json",
      "checker": {
        "product_id": "agent-audit-coverage",
        "price_gbp_excluding_vat": 79,
        "purchase_url": "/index-access?route=agentic_pack_audit_coverage_v1",
        "licence_scope": "Existing separate internal-use licence; subscriptions do not include the executable.",
        "benefit": "Optional local analysis of supported supplied declarations only; no native provider integration or verification is added."
      },
      "boundary": "Optional existing checker analyses its supported supplied declarations locally. It does not connect to the documented provider, import native provider exports, verify tokens, establish live outcomes or extend its schema or licence."
    },
    "exact_access": {
      "endpoint": "/mcp/catalogue",
      "tool": "get_publication_record",
      "arguments": {
        "record_id": "VXI-EXT-ICE09-OPENAI-OUTPUT-TRIPWIRE-EXCEPTION-SESSION-EVIDENCE-01-v1"
      },
      "record_url": "/varexis-index/records/openai-output-tripwire-exception-session-evidence.json",
      "original_mcp_search": false,
      "paid_saved_retrieval": false
    }
  },
  "niche": {
    "niche_id": "ice09-audit",
    "label": "Audit, Provenance & Evidence Accountability",
    "classification_id": "audit",
    "classification_label": "Audit, Provenance & Evidence Accountability",
    "topic_id": "audit",
    "buyer_intent": "Does an output guardrail tripwire leave the same session evidence as a guardrail exception? Do not infer absent tool work from withheld output, assume rejected payloads are retained, or treat a guardrail exception as a rejection verdict. Inspect the actual session write and retained fields.",
    "comparison_role": "supplementary_discovery"
  },
  "paid_access_note": "The full published answer and boundaries are free. Paid access saves the selected route and source versions, with an unsigned receipt when purchased. Preview coverage before paying. Fresh research, bespoke assessment and protected evidence are excluded.",
  "service_route": {
    "request_guide": "/varexis-index/request.html",
    "request_workflow": "/varexis-index/request-workflow.json",
    "service_id": "single_request",
    "service_routes_url": "/varexis-index/service-routes.json",
    "payment_required": true,
    "availability_endpoint": "/api/index/capabilities",
    "operational_status": "AUTOMATIC_SERVICE_CAPABILITIES_GATED",
    "checkout_url": "/index-access"
  },
  "authority_boundary": "VAREXIS-INDEX informs evidence reliance. It does not provide advice, certification, deployment authority, security assurance, production approval, or guarantee.",
  "human_page": "/varexis-index/catalog.html#openai-output-tripwire-exception-session-evidence",
  "machine_readable_record": "/varexis-index/records/openai-output-tripwire-exception-session-evidence.json"
}