Agent Run Forensics & Handoff Kit

Turn a failed or paused agent run into a redacted deterministic timeline, conservative retry review, and manual handoff artifacts.

Agent operators and developers handing off failed or paused runs without replaying tools or exposing submitted credentials.

Agent run packet (strict JSON)

{
  "run_id": "run-sample-1",
  "goal": "Hand off a failed checkout agent run.",
  "run_status": "failed",
  "events": [
    {
      "id": "event-1",
      "sequence": 1,
      "event_kind": "tool",
      "status": "complete",
      "tool_name": "lookup_order",
      "argument_keys": [
        "order_id"
      ],
      "side_effect_state": "none",
      "depends_on": [],
      "artifact_ids": [
        "artifact-1"
      ],
      "summary": "Order lookup completed."
    },
    {
      "id": "event-2",
      "sequence": 2,
      "event_kind": "tool",
      "status": "failed",
      "tool_name": "create_refund",
      "argument_keys": [
        "amount",
        "payment_intent"
      ],
      "error": {
        "code": "REFUND_FAILED",
        "message": "The submitted tool result reported failure."
      },
      "side_effect_state": "unknown",
      "depends_on": [
        "event-1"
      ],
      "artifact_ids": [],
      "summary": "Refund creation failed."
    }
  ],
  "decisions": [
    {
      "id": "decision-1",
      "summary": "Stop automatic work.",
      "evidence_ids": [
        "evidence-1"
      ]
    }
  ],
  "blockers": [
    {
      "id": "blocker-1",
      "summary": "Refund side effects are unknown.",
      "next_step": "Inspect the provider dashboard manually."
    }
  ],
  "next_actions": [
    {
      "id": "action-1",
      "summary": "Confirm refund state manually.",
      "priority": "high",
      "depends_on": []
    }
  ],
  "artifacts": [
    {
      "id": "artifact-1",
      "kind": "log",
      "ref": "local/redacted-run.log"
    }
  ],
  "evidence": [
    {
      "id": "evidence-1",
      "source_id": "submitted-run",
      "quote": "The caller marked side effects unknown."
    }
  ],
  "options": {
    "max_pack_bytes": 16384,
    "report_format": "markdown",
    "report_mode": "details"
  }
}

Generation stays in this browser tab and calls the same deterministic core directly. No tool is replayed, no retry is started, and no input is stored or sent over the network.

The visible sample was built locally during this static build. Downloads remain disabled until you build the current input.

Timeline

[
  {
    "id": "event-1",
    "ordinal": 1,
    "sequence": 1,
    "event_kind": "tool",
    "status": "complete",
    "tool_name": "lookup_order",
    "argument_keys": [
      "order_id"
    ],
    "error": null,
    "side_effect_state": "none",
    "depends_on": [],
    "artifact_ids": [
      "artifact-1"
    ],
    "summary": "Order lookup completed.",
    "submitted_unverified": true,
    "root_cause": null
  },
  {
    "id": "event-2",
    "ordinal": 2,
    "sequence": 2,
    "event_kind": "tool",
    "status": "failed",
    "tool_name": "create_refund",
    "argument_keys": [
      "amount",
      "payment_intent"
    ],
    "error": {
      "code": "REFUND_FAILED",
      "message": "The submitted tool result reported failure."
    },
    "side_effect_state": "unknown",
    "depends_on": [
      "event-1"
    ],
    "artifact_ids": [],
    "summary": "Refund creation failed.",
    "submitted_unverified": true,
    "root_cause": null
  }
]

Forensics and handoff envelope

{
  "ok": true,
  "schema_version": "utilito.product-kit.v1",
  "kit": {
    "id": "agent-run-forensics-handoff-kit",
    "version": "1.0.0"
  },
  "result": {
    "status": "review_required",
    "root_cause": null,
    "timeline": [
      {
        "id": "event-1",
        "ordinal": 1,
        "sequence": 1,
        "event_kind": "tool",
        "status": "complete",
        "tool_name": "lookup_order",
        "argument_keys": [
          "order_id"
        ],
        "error": null,
        "side_effect_state": "none",
        "depends_on": [],
        "artifact_ids": [
          "artifact-1"
        ],
        "summary": "Order lookup completed.",
        "submitted_unverified": true,
        "root_cause": null
      },
      {
        "id": "event-2",
        "ordinal": 2,
        "sequence": 2,
        "event_kind": "tool",
        "status": "failed",
        "tool_name": "create_refund",
        "argument_keys": [
          "amount",
          "payment_intent"
        ],
        "error": {
          "code": "REFUND_FAILED",
          "message": "The submitted tool result reported failure."
        },
        "side_effect_state": "unknown",
        "depends_on": [
          "event-1"
        ],
        "artifact_ids": [],
        "summary": "Refund creation failed.",
        "submitted_unverified": true,
        "root_cause": null
      }
    ],
    "timeline_pack": {
      "ok": true,
      "schema_version": "utilito.context-pack.v1",
      "recipe": {
        "id": "agent-run-timeline-pack",
        "version": "1.0.0"
      },
      "format": "review",
      "output": "{\"schema_version\":\"utilito.agent-run-timeline-pack.v1\",\"review_id\":\"run-sample-1\",\"run_status\":\"failed\",\"summary\":{\"events\":2,\"complete\":1,\"failed\":1,\"blocked\":0,\"started\":0,\"skipped\":0},\"records\":[{\"id\":\"timeline-1\",\"status\":\"event\",\"ordinal\":1,\"event_kind\":\"tool\",\"outcome\":\"complete\",\"tool_name\":\"lookup_order\",\"argument_keys\":[\"order_id\"],\"error_code\":null},{\"id\":\"timeline-2\",\"status\":\"event\",\"ordinal\":2,\"event_kind\":\"tool\",\"outcome\":\"failed\",\"tool_name\":\"create_refund\",\"argument_keys\":[\"amount\",\"payment_intent\"],\"error_code\":\"REFUND_FAILED\"}],\"notes\":[],\"notice\":\"Submitted event order and labels only; no root cause or event truth was inferred.\",\"omissions\":{\"event\":0}}",
      "evidence": [],
      "receipt": {
        "input_bytes": 651,
        "output_bytes": 680,
        "evidence_bytes": 2,
        "bounded_payload_bytes": 682,
        "estimated_tokens_before": 163,
        "estimated_tokens_after": 170,
        "max_payload_bytes": 16384,
        "max_output_bytes": 16384,
        "max_input_records": 10000,
        "input_records": 2,
        "output_records": 2,
        "omitted_records": 0,
        "omitted_fields": 0,
        "deduplicated": 0,
        "truncated": false,
        "review_schema_version": "utilito.agent-run-timeline.v1",
        "output_schema_version": "utilito.agent-run-timeline-pack.v1",
        "selected_by_status": {
          "event": 2
        },
        "omitted_by_status": {
          "event": 0
        },
        "content_hash": "abe3c531",
        "warnings": []
      },
      "provenance": {
        "deterministic": true,
        "model_used": false,
        "network_used": false,
        "storage_used": false,
        "core": "utilito-context-pack",
        "core_version": "1.0.0"
      },
      "error": null
    },
    "failure_pack": {
      "ok": true,
      "schema_version": "utilito.context-pack.v1",
      "recipe": {
        "id": "agent-run-failure-pack",
        "version": "1.0.0"
      },
      "format": "review",
      "output": "{\"schema_version\":\"utilito.agent-run-failure-review.v1\",\"run_id\":\"run-sample-1\",\"goal\":\"Hand off a failed checkout agent run.\",\"status\":\"failed\",\"steps\":[{\"id\":\"event-1\",\"tool_id\":\"lookup_order\",\"status\":\"complete\",\"argument_keys\":[\"order_id\"]},{\"id\":\"event-2\",\"tool_id\":\"create_refund\",\"status\":\"failed\",\"argument_keys\":[\"amount\",\"payment_intent\"],\"error\":{\"code\":\"REFUND_FAILED\",\"message\":\"The submitted tool result reported failure.\"}}],\"notes\":[],\"notice\":\"Submitted run state is packaged only; it is not replayed and no root cause is inferred.\",\"omissions\":{\"failed\":0,\"blocked\":0,\"skipped\":0,\"complete\":0}}",
      "evidence": [
        {
          "id": "evidence-1",
          "source_id": "submitted-run",
          "quote": "The caller marked side effects unknown."
        }
      ],
      "receipt": {
        "max_payload_bytes": 16384,
        "input_bytes": 555,
        "output_bytes": 612,
        "evidence_bytes": 99,
        "bounded_payload_bytes": 711,
        "estimated_tokens_before": 139,
        "estimated_tokens_after": 153,
        "input_records": 2,
        "output_records": 2,
        "omitted_records": 0,
        "evidence_input_records": 1,
        "evidence_output_records": 1,
        "evidence_omitted_records": 0,
        "omitted_fields": 0,
        "deduplicated": 0,
        "truncated": false,
        "review_schema_version": "utilito.agent-run-failure.v1",
        "selected_by_status": {
          "failed": 1,
          "blocked": 0,
          "skipped": 0,
          "complete": 1
        },
        "omitted_by_status": {
          "failed": 0,
          "blocked": 0,
          "skipped": 0,
          "complete": 0
        },
        "content_hash": "020a5fef",
        "warnings": []
      },
      "provenance": {
        "deterministic": true,
        "model_used": false,
        "network_used": false,
        "storage_used": false,
        "core": "utilito-context-pack",
        "core_version": "1.0.0"
      },
      "error": null
    },
    "failure_pack_reason": null,
    "redacted_handoff_pack": {
      "ok": true,
      "schema_version": "utilito.context-pack.v1",
      "recipe": {
        "id": "agent-handoff-pack",
        "version": "1.0.0"
      },
      "format": "handoff",
      "output": "{\"schema_version\":\"utilito.agent-handoff.v1\",\"handoff_id\":\"handoff:run-sample-1\",\"goal\":\"Hand off a failed checkout agent run.\",\"status\":\"blocked\",\"decisions\":[{\"evidence_ids\":[\"evidence:evidence-1\"],\"id\":\"decision:decision-1\",\"summary\":\"Stop automatic work.\"}],\"completed\":[{\"artifact_ids\":[\"artifact:artifact-1\"],\"id\":\"completed:event-1\",\"summary\":\"Order lookup completed.\"}],\"blockers\":[{\"id\":\"blocker:blocker-1\",\"next_step\":\"Inspect the provider dashboard manually.\",\"summary\":\"Refund side effects are unknown.\"}],\"next_actions\":[{\"depends_on\":[],\"id\":\"action:action-1\",\"priority\":\"high\",\"summary\":\"Confirm refund state manually.\"}],\"artifacts\":[{\"id\":\"artifact:artifact-1\",\"kind\":\"log\",\"ref\":\"local/redacted-run.log\"}],\"evidence\":[{\"id\":\"evidence:evidence-1\",\"quote\":\"The caller marked side effects unknown.\",\"source_id\":\"submitted-run\"}],\"unknowns\":[],\"constraints\":[],\"notes\":[],\"omissions\":{\"decisions\":0,\"completed\":0,\"blockers\":0,\"next_actions\":0,\"artifacts\":0,\"evidence\":0,\"unknowns\":0,\"constraints\":0,\"notes\":0}}",
      "evidence": [
        {
          "id": "evidence:evidence-1",
          "quote": "The caller marked side effects unknown.",
          "source_id": "submitted-run"
        }
      ],
      "receipt": {
        "input_bytes": 855,
        "output_bytes": 1024,
        "evidence_bytes": 108,
        "bounded_payload_bytes": 1132,
        "estimated_tokens_before": 214,
        "estimated_tokens_after": 256,
        "max_payload_bytes": 16384,
        "max_output_bytes": 16384,
        "max_input_records": 10000,
        "input_records": 6,
        "output_records": 6,
        "omitted_records": 0,
        "omitted_fields": 0,
        "deduplicated": 0,
        "truncated": false,
        "handoff_schema_version": "utilito.agent-handoff.v1",
        "selected_by_section": {
          "decisions": 1,
          "completed": 1,
          "blockers": 1,
          "next_actions": 1,
          "artifacts": 1,
          "evidence": 1,
          "unknowns": 0,
          "constraints": 0,
          "notes": 0
        },
        "omitted_by_section": {
          "decisions": 0,
          "completed": 0,
          "blockers": 0,
          "next_actions": 0,
          "artifacts": 0,
          "evidence": 0,
          "unknowns": 0,
          "constraints": 0,
          "notes": 0
        },
        "reference_count": 2,
        "unresolved_references": 0,
        "warnings": []
      },
      "provenance": {
        "deterministic": true,
        "model_used": false,
        "network_used": false,
        "storage_used": false,
        "core": "utilito-context-pack",
        "core_version": "1.0.0"
      },
      "error": null
    },
    "retry_safety": {
      "classification": "unknown",
      "basis": "caller_submitted_side_effect_state_unverified",
      "event_ids": [
        "event-2"
      ],
      "automatic_retry": false
    },
    "reports": {
      "handoff": {
        "ok": true,
        "schema_version": "utilito.evidence-report.v1",
        "preset": {
          "id": "agent-handoff-report",
          "version": "1.0.0"
        },
        "export_format": "markdown",
        "mime_type": "text/markdown",
        "content": "# Agent run handoff\n\n> This report preserves submitted handoff state and transformation receipts. It does not verify that the submitted claims are true.\n\n- Recipe: `[email protected]`\n- Source contract: `utilito.agent-handoff.v1`\n- Storage: response-only; not stored by Artifact Ops\n\n## Goal\n\n> Hand off a failed checkout agent run\\.\n\n## Status\n\n`blocked`\n\n## Decisions\n\n### decision:decision\\-1\n\nStop automatic work\\.\n\n- Evidence IDs: `evidence:evidence\\-1`\n\n## Completed\n\n### completed:event\\-1\n\nOrder lookup completed\\.\n\n- Artifact IDs: `artifact:artifact\\-1`\n\n## Blockers\n\n### blocker:blocker\\-1\n\nRefund side effects are unknown\\.\n\n- Next step: Inspect the provider dashboard manually\\.\n\n## Next actions\n\n### action:action\\-1\n\nConfirm refund state manually\\.\n\n- Priority: high\n\n## Artifacts\n\n### artifact:artifact\\-1\n\n\n\n- Kind: log\n- Reference: local/redacted\\-run\\.log\n\n## Evidence\n\n### evidence:evidence\\-1\n\nThe caller marked side effects unknown\\.\n\n- Source ID: submitted\\-run\n\n## Unknowns\n\nNo retained items.\n\n## Constraints\n\nNo retained items.\n\n## Notes\n\nNo retained notes.\n\n## Omissions\n\n| Section | Omitted |\n|---|---:|\n| decisions | 0 |\n| completed | 0 |\n| blockers | 0 |\n| next\\_actions | 0 |\n| artifacts | 0 |\n| evidence | 0 |\n| unknowns | 0 |\n| constraints | 0 |\n| notes | 0 |\n\n## Transformation receipt\n\n| Metric | Value |\n|---|---:|\n| Input bytes | 855 |\n| Output bytes | 1024 |\n| Estimated tokens before | 214 |\n| Estimated tokens after | 256 |\n| Input records | 6 |\n| Output records | 6 |\n| Omitted records | 0 |\n| Omitted fields | 0 |\n| Deduplicated | 0 |\n| Truncated | no |\n\n",
        "receipt": {
          "input_bytes": 2329,
          "output_bytes": 1603,
          "sections": 10,
          "evidence_items": 1,
          "truncated": false,
          "sanitized_html_fields": 0
        },
        "provenance": {
          "source_schema_version": "utilito.context-pack.v1",
          "deterministic": true,
          "model_used": false,
          "network_used": false,
          "storage_used": false,
          "core": "utilito-evidence-report",
          "core_version": "1.0.0"
        },
        "error": null
      },
      "failure": {
        "ok": true,
        "schema_version": "utilito.evidence-report.v1",
        "preset": {
          "id": "agent-run-failure-report",
          "version": "1.0.0"
        },
        "export_format": "markdown",
        "mime_type": "text/markdown",
        "content": "# Agent run failure review\n\n> Submitted run state is packaged only; it is not replayed and no root cause is inferred\\.\n\n## Review identity\n\n- Run: run\\-sample\\-1\n- Goal: Hand off a failed checkout agent run\\.\n- Status: failed\n\n## Run steps\n\n### event\\-1\n\n- Tool: lookup\\_order\n- Status: complete\n- Argument keys: `order\\_id`\n\n### event\\-2\n\n- Tool: create\\_refund\n- Status: failed\n- Argument keys: `amount`, `payment\\_intent`\n- Error code: REFUND\\_FAILED\n- Error: The submitted tool result reported failure\\.\n\n## Omissions\n\n- failed: 0\n- blocked: 0\n- skipped: 0\n- complete: 0\n\n## Transformation receipt\n\n- Input records: 2\n- Output records: 2\n- Omitted records: 0\n- Payload bytes: 711/16384\n\n",
        "receipt": {
          "input_bytes": 1720,
          "output_bytes": 691,
          "sections": 4,
          "evidence_items": 1,
          "truncated": false,
          "sanitized_html_fields": 0
        },
        "provenance": {
          "source_schema_version": "utilito.context-pack.v1",
          "deterministic": true,
          "model_used": false,
          "network_used": false,
          "storage_used": false,
          "core": "utilito-evidence-report",
          "core_version": "1.0.0"
        },
        "error": null
      }
    },
    "redaction_receipt": {
      "fields": 0,
      "replacements": 0,
      "marker": "[REDACTED]"
    },
    "safety": {
      "tools_replayed": false,
      "tools_executed": false,
      "automatic_retry": false,
      "root_cause_inferred": false,
      "input_stored": false,
      "network_used": false
    },
    "safe_actions": [
      {
        "code": "REVIEW_TIMELINE",
        "message": "Review submitted event order and unverified labels manually.",
        "automatic": false
      },
      {
        "code": "INSPECT_SIDE_EFFECTS",
        "message": "Inspect side effects before any manual retry.",
        "automatic": false
      },
      {
        "code": "HAND_OFF_MANUALLY",
        "message": "Share only the redacted handoff artifacts with the next operator.",
        "automatic": false
      }
    ]
  },
  "receipt": {
    "input_bytes": 1294,
    "output_bytes": 10857,
    "operations": 4,
    "omitted_items": 0,
    "warnings": [],
    "events": 2,
    "redacted_fields": 0,
    "redaction_replacements": 0
  },
  "provenance": {
    "deterministic": true,
    "model_used": false,
    "network_used": false,
    "storage_used": false,
    "core_versions": {
      "agent_run_forensics_kit": "1.0.0",
      "context_pack": "1.0.0",
      "deterministic_review_pack": "1.0.0",
      "product_review_pack": "1.0.0",
      "evidence_report": "1.0.0"
    }
  },
  "error": null
}

Download handoff pack JSON Download timeline pack JSON Download failure pack JSON Download handoff report Download failure report

Free runnable capability

Generate one bounded local forensics and handoff packet with deterministic redaction and no signup.

Builder boundary

The existing Builder Pilot can coordinate higher bounded review volume; it does not add replay, execution, automatic retry, storage, or root-cause inference.

See the existing Builder Pilot

Local and response-only privacy

Browser generation runs locally with the shared deterministic cores. API and MCP requests are response-only: no storage, telemetry, network calls, model calls, replay, execution, or automatic retry.

Limits and trust boundary

All statuses, summaries, side-effect states, decisions, and evidence are caller-submitted unverified labels. The kit never infers root cause or truth.
Only tool names and argument key names are accepted. Argument values, tool outputs, replay, execution, network access, storage, telemetry, and model calls are outside the contract.
Retry classification is conservative metadata only: possible or confirmed side effects are unsafe, unknown stays unknown, and safe requires every failed or blocked event to be explicitly none. No retry is automatic.
Credential-like text is replaced with [REDACTED] before any pack or report is built. Review the generated artifacts manually before sharing.