Sigil: Complexity Example Ladder

• the user asks to explain, illustrate, or provide examples,
• a caller contract requires complexity-calibrated examples,
• multiple options need comparable examples before a human choice,
• interacting identity, state, dependency, boundary, or failure conditions make an abstract distinction hard to evaluate.

Do not invoke it for unrelated factual or mechanical answers merely to add prose. For Decision Gate, invoke only after the user selects "Explain / more context" or otherwise signals uncertainty.

• the subject, question, or decision being explained,
• authoritative source evidence and its claim ceiling,
• the invariant concept or distinction that must stay fixed across all rungs,
• every admissible action option when the mode is comparative,
• defer and stop choices when their continuation effects matter,
• constraints, assumptions, and authority boundaries,
• caller and trigger kind,
• audience baseline and optional complexity axes.

If the source cannot support a rung and a safe hypothetical cannot be labeled, return BLOCK rather than fabricate the example.

These are structural differences, not word-count, jargon, or detail targets.

When repository observability is available, record the standard sigil signal plus:

• trigger kind and caller,
• explanatory or comparative mode,
• required and produced rungs,
• complexity axes,
• invariant-preservation result,
• required options and per-rung coverage,
• hypothetical count,
• unsupported-claim count,
• unequal-option-coverage flag,
• decision effect,
• user correction signal.

Default reflection triggers are 5 meaningful executions, 10 generated outputs, 3 related gaps, or 1 severe gap. Severe gaps include unlabelled fabrication, examples treated as consent or selection, material option bias, and private/public or authority-boundary leakage.

• produce low, medium, and complex rungs,
• increase structural complexity at every rung,
• preserve one stable concept or comparison invariant,
• distinguish evidence from labelled hypothetical details,
• use one shared scenario per rung in comparative mode,
• cover every admissible action option evenly,
• keep defer and stop as continuation states rather than invented product behavior,
• state what each example demonstrates and does not prove,
• report any unsupported claim instead of hiding it,
• preserve the caller's recommendation, admissibility, gate, and authority state,
• keep public templates and experiment fixtures product-neutral,
• return decision effect: none whenever a decision is involved.

• increasing length without increasing structural complexity,
• changing the core subject or invariant between rungs,
• equating jargon with complexity,
• introducing unsupported domain facts,
• presenting a hypothetical as source evidence,
• giving a recommended option richer or more favorable examples,
• using unrelated scenarios that prevent controlled comparison,
• silently omitting a rung or admissible action option,
• turning an example into a recommendation, choice, consent, or owner approval,
• treating the ladder as application-specific validity or admissibility proof,
• expanding into implementation, mutation, promotion, publication, or execution,
• copying caller-private material into public sigil examples or fixtures.

## Complexity Example Ladder Result

- Status: pass | block
- Subject: <subject or decision>
- Mode: explanatory | comparative
- Complexity basis: <structural axes>
- Evidence: <sources or supplied context>
- Assumptions: <labelled hypotheticals or none>

### Low
- Scenario: <shared low scenario>
- Example: <concept behavior or every option under the scenario>
- Demonstrates: <what becomes visible>
- Limits: <what this does not prove>

### Medium
- Added complexity: <structural addition>
- Scenario: <shared medium scenario>
- Example: <concept behavior or every option under the scenario>
- Demonstrates: <what becomes visible>
- Limits: <what this does not prove>

### Complex
- Added complexity: <structural addition>
- Scenario: <shared complex scenario>
- Example: <concept behavior or every option under the scenario>
- Demonstrates: <what becomes visible>
- Limits: <what this does not prove>

### Coverage
- Rungs produced: <n>/3
- Options covered: <IDs or not applicable>
- Unsupported claims: <count and details>
- Decision effect: none | not applicable
- Next step: <return to caller | supply missing evidence>