skills/desloppify/SKILL.md
Improve code quality in a repository using desloppify. Use when auditing a codebase, raising code quality scores, cleaning up maintainability issues, or systematically working through desloppify findings.
npx skillsauth add agmangas/agent-skills desloppifyInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Use this skill to improve the quality of the codebase with desloppify.
Raise the strict score as high as possible through real code improvements. Do not game the score or apply cosmetic fixes that do not meaningfully improve the code.
$ARGUMENTS is the scan target...Set:
SCAN_PATH="${ARGUMENTS:-.}"
This workflow requires Python 3.11+.
Install and update the skill guidance:
pip install --upgrade "desloppify[full]"
desloppify update-skill claude
If the environment is not using Claude, use the appropriate target instead: claude, cursor, codex, copilot, windsurf, or gemini.
Before scanning, inspect the repository for directories that should obviously be excluded, such as:
Exclude obvious candidates with:
desloppify exclude <path>
If an exclusion is questionable, stop and ask for confirmation before applying it.
Run:
desloppify scan --path "$SCAN_PATH"
desloppify next
This is your main job. Repeat the following until blocked, complete, or awaiting confirmation:
desloppify next.desloppify.desloppify next again.Do not optimize for closing tickets quickly. Optimize for durable code quality improvements.
Apply the same care to both large refactors and small detailed fixes.
Use:
desloppify plan
when it helps reorder priorities or cluster related work into a more efficient sequence.
Rescan periodically as the codebase changes:
desloppify scan --path "$SCAN_PATH"
desloppify scan and desloppify next.development
Use when the user wants an AI coding agent to offload suitable low-risk, bounded codebase browsing, inventory, extraction, log triage, or simple single-file reasoning tasks to a local LM Studio model while keeping high-level reasoning and final decisions in the main model.
development
Use when the user explicitly asks for plain language, less jargon, a concise explanation, mentor-style codebase guidance, or an explanation for a developer who knows software engineering but is new to the project or domain.
development
Use when preparing files for NotebookLM, organizing documents into a knowledge base, converting formats for NotebookLM compatibility, or reducing a large document collection to fit NotebookLM's 50-source limit. Scores and prioritizes sources, performs strategic merging (time-series, topic-based, format consolidation), converts unsupported formats (PPTX to PDF, XLSX to CSV), applies flat structure with descriptive snake_case names, and optimizes for RAG retrieval performance.
tools
Analyze git history for commit style, stage changes logically, and commit without pushing. Use when the user wants to commit changes matching their repository's existing style.