Sigil: Skill Transcriptor

• source file or source description,
• desired target name,
• known target tier,
• local examples of successful use,
• source validation rules,
• source templates,
• desired output folder.

<tier>/<target-name>/

If the user has not approved file creation, return a package plan and conversion report instead of creating files.

1. Identify the source artifact and target repository.
2. Detect whether the source is a skill, prompt, command, workflow, or mixed document.
3. Capture the requested target name or derive a neutral candidate name.
4. Record source assumptions and missing context.

Step 2 - Coherence Gate

5. Decide whether the source is one coherent reusable capability.
6. If the source contains multiple separable capabilities, stop direct conversion and recommend skill-decomposer.
7. If the source is coherent but has unclear boundaries, return flag with required clarification.
8. If the source is coherent and reusable, continue.

Step 3 - Extract Behavior

9. Extract objective, triggers, modes, inputs, required context, process steps, gates, outputs, validation, tools, and observability.
10. Separate reusable behavior from source-specific vocabulary, local routing, private artifact names, and implementation details.
11. Preserve important gates even when source wording changes.

Step 4 - Classify Tier

12. Classify the target as Formulae, Transmutations, or Arcana.
13. Record tier rationale and rejected tiers.
14. Escalate to skill-decomposer if tier ambiguity comes from multiple capabilities in the source.

Step 5 - Plan Package

15. Define target folder, README.md, SKILL.md, templates, examples, registry updates, pack updates, and roadmap updates.
16. Draft neutral content guidance for each artifact.
17. Identify whether templates are required or optional.
18. Identify whether observability should use the general sigil hook, sigil-local telemetry, or no telemetry.

Step 6 - Validate Conversion

19. Check Arcanum entry requirements: dedicated folder, human README, executable skill contract, Quality Bar, Anti-Patterns, output contract, and observability guidance.
20. Check neutral wording and coupling risks.
21. Check links, JSON templates, and registry references when files exist.
22. Return pass, flag, or block.

• reject direct conversion when the source contains multiple separable capabilities,
• preserve reusable behavior while removing source-specific machinery,
• classify the target tier with rationale,
• define README, SKILL, template, registry, and validation expectations,
• include Quality Bar, Anti-Patterns, output contract, and observability guidance,
• route unclear boundaries to skill-decomposer,
• distinguish source facts, conversion inference, and authored target content.

• converting a tangled source as one large sigil,
• copying source frontmatter without checking compatibility,
• preserving private routing or local workflow names as canonical Arcanum vocabulary,
• treating tool permissions as portable without review,
• creating templates that do not support the output contract,
• registering the sigil before validation,
• replacing sigil-development lifecycle governance.

• source type,
• target name,
• target tier,
• conversion accepted or rejected,
• decomposition recommended,
• coupling risks found,
• files planned,
• files created,
• validation result,
• blocker count.

## Skill Transcriptor Result

- Source: <path or description>
- Mode: assess | convert | validate | register-plan | observe
- Target name: <name>
- Target tier: formulae | transmutations | arcana | undecided
- Direct conversion: pass | flag | block
- Decomposition recommended: yes | no
- Coupling risks: <count>
- Files planned: <paths>
- Files created: <paths or none>
- Registry updates: <planned or none>
- Validation: pass | flag | block | not run
- Next action: <action>