Desloppify-style code health auditing skill for Codex.
It analyzes an existing repository with grouped subagents, then validates, deduplicates, scores, and prioritizes findings.

This repository packages a Codex skill named desloppify-code-health-auditor with:

• A strict read-only auditing contract (no builds, no scaffolding, no file edits unless explicitly requested)
• Detector coverage for dead code, duplication, complexity, dependency coupling, naming consistency, and debug/log leftovers
• Grouped, non-clobbering subagent planning
• Deterministic findings validation and merge/score reporting

Core skill definition:

• desloppify-code-health-auditor/SKILL.md

• Produces a consistent audit report format across runs
• Reduces subagent collisions with grouped execution and unique output paths
• Normalizes severity/tier/confidence across heterogeneous detector outputs
• Generates actionable next steps (top findings, remediation plan, resolve simulation commands)
• Works with plain Python scripts and JSON artifacts

desloppify-code-health-auditor/
├── SKILL.md
├── agents/openai.yaml
├── references/grouped-subagent-prompts.md
└── scripts/
    ├── repo_profile.py
    ├── grouped_subagent_plan.py
    ├── validate_findings.py
    └── merge_and_score.py

• Python 3.9+
• A repository to audit
• A Codex environment that can run subagents (for full workflow)

Clone this repository and place the skill folder where your Codex instance loads skills (commonly $CODEX_HOME/skills):

git clone [email protected]:davelindo/codex-skill-deslopify.git
cd codex-skill-deslopify

Then make sure the desloppify-code-health-auditor/ directory is available in your Codex skills path.

From the skill directory:

cd desloppify-code-health-auditor

1. Profile the target repository:

python3 scripts/repo_profile.py /path/to/target-repo --out runs/profile.json

2. Generate grouped, non-clobbering tasks:

python3 scripts/grouped_subagent_plan.py \
  --profile runs/profile.json \
  --run-id run-001 \
  --out runs/run-001/plan.json

3. Execute subagents per plan and write JSON arrays to:

• runs/run-001/findings/*.json

4. Validate subagent outputs:

python3 scripts/validate_findings.py runs/run-001/findings

5. Merge/deduplicate/score results:

python3 scripts/merge_and_score.py \
  runs/run-001/findings \
  --out runs/run-001/report.json

Use $desloppify-code-health-auditor to audit this repository and return a deduplicated, prioritized findings report.

Prompt templates and detector snippets:

• desloppify-code-health-auditor/references/grouped-subagent-prompts.md

• Open an issue in this repository: https://github.com/davelindo/codex-skill-deslopify/issues
• Review the skill contract: desloppify-code-health-auditor/SKILL.md
• Review prompt templates: desloppify-code-health-auditor/references/grouped-subagent-prompts.md

Maintainer:

• @davelindo

Contributions are welcome via pull requests. To keep changes easy to review:

1. Keep edits focused (skill spec, scripts, or references).
2. Include a short rationale in the PR description.
3. Run the script workflow locally on sample findings before opening the PR.
4. Avoid committing generated runs/ artifacts.