incoherence
Detect and resolve incoherence in documentation, code, specs vs implementation. Includes reconciliation phase for applying user-provided resolutions.
When & Why to Use This Skill
The Incoherence Detector is a powerful Claude skill designed to automate the identification and resolution of discrepancies between technical documentation, source code, and project specifications. By utilizing a structured 22-step workflow, it ensures that software implementations remain strictly aligned with their design requirements, significantly reducing technical debt and documentation rot in complex development environments.
Use Cases
- Technical Debt Reduction: Identifying and fixing outdated README files or API documentation that no longer matches the current codebase implementation.
- Specification Validation: Verifying that new feature code strictly adheres to the logic and constraints defined in initial technical design documents (TDD).
- Cross-Artifact Consistency: Detecting naming or logic mismatches between architectural diagrams, inline comments, and the actual executable code.
- Guided Reconciliation: Using the multi-step workflow to systematically apply user-approved fixes to documentation and code after a major refactoring session.
| name | incoherence |
|---|---|
| description | Detect and resolve incoherence in documentation, code, specs vs implementation. Includes reconciliation phase for applying user-provided resolutions. |
Incoherence Detector
When this skill activates, IMMEDIATELY invoke the script. The script IS the workflow.
Prerequisites
User must specify the report filename (e.g., "output to incoherence-report.md").
Invocation
# Detection phase (steps 1-13)
python3 scripts/incoherence.py \
--step-number 1 \
--total-steps 22 \
--thoughts "<context>"
# Reconciliation phase (steps 14-22, after user edits report)
python3 scripts/incoherence.py \
--step-number 14 \
--total-steps 22 \
--thoughts "Reconciling user resolutions from report"
| Argument | Required | Description |
|---|---|---|
--step-number |
Yes | Current step (1-22) |
--total-steps |
Yes | Always 22 |
--thoughts |
Yes | Accumulated state from all previous steps |
Do NOT explore or detect first. Run the script and follow its output.