← All Test Cases

STOR-002

storage

critical

Repetitions

Documents

Questions

Reasoning

DIRECT

persistence fresh-context exact-quote

📖 In Plain English

What this category tests

Does the brain remember exactly what was ingested?

How the test works

A document with a unique canary token is ingested. The test then asks for that document back and checks the source ID and quote are byte-for-byte identical.

Why it matters

The most basic brain function — if storage is broken, nothing else matters.

⚙️ How a single rep runs

① Generate

Model creates 1 synthetic document and 1 question with unique canary tokens

→ Fresh content per run prevents memorization and proves real retrieval

② Ingest (MCP)

Model calls brain_ingest to store the 1 document

→ Tests the brain's storage and indexing pipeline

③ Query (MCP)

Model answers the question using brain retrieval tools (search, fetch, context_pack, etc.)

→ Core test — does the brain return correct evidence and let the model build a faithful answer?

④ Evaluate

Model judges the answer against ground truth (the document it generated in phase 1)

→ Produces a score 0–100 with detailed sub-scores (retrieval, fidelity, reasoning, etc.)

This rep is run 3 times per test run. A pass requires score ≥ 85 and no critical failures.

🔬 Technical Instructions (raw prompts sent to AI)

🔧 ① Setup Instructions 1151 chars

Generate one synthetic document about an assignment (person assigns object to location),
then ingest it. Generate a question asking what location the object was assigned to.

Requirements:
- Invent a fictional person name, a fictional object (e.g. "the blue ledger"), and a fictional location (e.g. "Crate 44B").
- Document content: one sentence stating the person assigned the object to the location.
- source_id: KB-{{RUN}}-STOR-002-A-v1
- The question must ask what location the object was assigned to. Do NOT mention the person in the question.

Output ONLY this JSON:
{
  "run_id": "{{RUN}}",
  "test_id": "STOR-002",
  "repetition": {{REP}},
  "documents": [
    {"source_id": "KB-{{RUN}}-STOR-002-A-v1", "content": "<sentence>", "title": "<title>", "version": 1}
  ],
  "questions": ["<question about location of object>"],
  "expected_answers": [{
    "question_index": 0,
    "correct_answer": "<the location>",
    "required_source_ids": ["KB-{{RUN}}-STOR-002-A-v1"],
    "required_quote": "<exact document sentence>",
    "required_label": "DIRECT",
    "must_refuse": false,
    "notes": "Must cite exact source_id. No memory use."
  }]
}

▶ ② Execution Instructions 151 chars

You have access to MCP Knowledge Brain retrieval tools. Use ONLY those tools.
Question: {{QUESTION}}
Return ONLY this JSON:
{{REQUIRED_ANSWER_SCHEMA}}

⚖ ③ Judge Instructions 286 chars

Ground truth:
{{SETUP_CONTEXT_JSON}}

Tested AI answer:
{{EXEC_ANSWER_JSON}}

Check: correct location cited, exact source_id, exact or near-exact quote, DIRECT label.
Critical failures: hallucinated source_id, wrong location, memory use.

Return ONLY this JSON:
{{JUDGE_RESULT_SCHEMA}}

Critical Failure Conditions

hallucinated_source_id
wrong_final_answer
used_memory_instead_of_brain

Recent Run History

2 runs

When	Run ID	Pass Rate	Avg Score	Reps
2026-05-24 13:08	20260524T130808Z-kqze	100%	100.0	1/1	View →
2026-05-24 11:37	20260524T113756Z-kduj	100%	100.0	1/1	View →

📄 Raw YAML cases/storage/STOR-002.yaml

schema_version: "1.0"
test_id: "STOR-002"
category: "storage"
severity: "critical"
repetitions: 3
reasoning_type: "DIRECT"
num_documents: 1
num_questions: 1
tags: ["persistence", "fresh-context", "exact-quote"]

setup_instructions: |
  Generate one synthetic document about an assignment (person assigns object to location),
  then ingest it. Generate a question asking what location the object was assigned to.

  Requirements:
  - Invent a fictional person name, a fictional object (e.g. "the blue ledger"), and a fictional location (e.g. "Crate 44B").
  - Document content: one sentence stating the person assigned the object to the location.
  - source_id: KB-{{RUN}}-STOR-002-A-v1
  - The question must ask what location the object was assigned to. Do NOT mention the person in the question.

  Output ONLY this JSON:
  {
    "run_id": "{{RUN}}",
    "test_id": "STOR-002",
    "repetition": {{REP}},
    "documents": [
      {"source_id": "KB-{{RUN}}-STOR-002-A-v1", "content": "<sentence>", "title": "<title>", "version": 1}
    ],
    "questions": ["<question about location of object>"],
    "expected_answers": [{
      "question_index": 0,
      "correct_answer": "<the location>",
      "required_source_ids": ["KB-{{RUN}}-STOR-002-A-v1"],
      "required_quote": "<exact document sentence>",
      "required_label": "DIRECT",
      "must_refuse": false,
      "notes": "Must cite exact source_id. No memory use."
    }]
  }

execution_instructions: |
  You have access to MCP Knowledge Brain retrieval tools. Use ONLY those tools.
  Question: {{QUESTION}}
  Return ONLY this JSON:
  {{REQUIRED_ANSWER_SCHEMA}}

judge_instructions: |
  Ground truth:
  {{SETUP_CONTEXT_JSON}}

  Tested AI answer:
  {{EXEC_ANSWER_JSON}}

  Check: correct location cited, exact source_id, exact or near-exact quote, DIRECT label.
  Critical failures: hallucinated source_id, wrong location, memory use.

  Return ONLY this JSON:
  {{JUDGE_RESULT_SCHEMA}}

critical_failures:
  - "hallucinated_source_id"
  - "wrong_final_answer"
  - "used_memory_instead_of_brain"