Tinder for your Claude Code skills. Reviews a sorted deck of every installed skill and lets you swipe keep / delete / skip on each one. Use when the user wants to bulk-clean their skill collection, triage unused skills, or do interactive skill cleanup. Trigger with '/janitor-swipe'.
npx skills add https://github.com/khendzel/skills-janitor --skill janitor-swipe
A bash TUI that puts every installed skill into a sorted deck and lets the user swipe keep / delete / skip on each card.
The deck is sorted "most likely waste first" — heavy, never-used skills appear at the top, so most users hit ← delete a few times and quit before reviewing the whole list. The swipe is scope-aware and honest about what it can delete: user-scope skills are actually removed (after confirmation), plugin skills are flagged for review instead.
scripts/swipe.sh and swipe-build-deck.sh)!Inside Claude Code, the Bash tool's stdin is non-interactive, so the keypress reader can't work. The user must invoke it via the ! prefix so the command runs in their actual shell:
!bash ~/.claude/skills/skills-janitor/scripts/swipe.sh
When the user asks for /janitor-swipe, tell them to run that command in their terminal. Do NOT try to run it yourself via the Bash tool — it will error with "needs an interactive terminal".
Each card shows:
[3 / 47])X always · Y on trigger (description tokens are permanent; body loads only when the skill fires)user, project, plugin · <plugin-name>, etc.)Controls:
← / h / d — stage for delete→ / l / k — keep↓ / j / s / space — skipu — undo (back up one card, clear its decision)i — inspect (show full SKILL.md description)q / Esc — quit (still shows summary for decisions made so far)| Scope | What happens on swipe left |
|---|---|
| user, project, codex-user, codex-project | Path is staged for rm -rf (or unlink if symlink) |
| plugin, plugin-source | NOT deleted — flagged under "Plugins to review" at the apply screen, with a hint to run /plugin uninstall <plugin> if enough skills from that plugin were swiped |
| mcp-user, mcp-project (v1.7) | Server entry is removed from its config file (~/.claude.json / .mcp.json) with a timestamped .bak backup — reversible |
| mcp-plugin | NOT edited — flagged for plugin review like plugin skills |
MCP server cards show real call counts from session transcripts instead of token guesses (schemas live server-side). An unused connected server ranks high — its tool schemas load into context for nothing.
After the last card (or q), an apply screen shows keep/skip/delete counts, the deletion list with paths, the "frees X always-loaded + Y on-trigger" token summary, the plugin review breakdown, and a prompt:
y — apply deletions immediately (logged to ~/.skills-janitor/log.jsonl with path and frontmatter snapshot)N — cancelsave — write decisions to ~/.skills-janitor/swipe-<timestamp>.json for later application via swipe.sh --apply <file>Solution: The command was run through the Bash tool. Have the user run it with the ! prefix in their own terminal.
Solution: The terminal window is too small — resize and retry.
Solution: Nothing is installed (exits 0 with a message) — suggest /janitor-discover to find skills.
Solution: Point them at /janitor-report or /janitor-value for the same data in list form.
Edge cases handled by the script: Ctrl-C mid-swipe restores the terminal; symlinks are unlinked, never followed; old saved decks without the v1.5 token-split fields still load.
Input: "/janitor-swipe" or "help me clean up my skills interactively"
Output: Explain the flow in two sentences, then give the exact command to run: !bash ~/.claude/skills/skills-janitor/scripts/swipe.sh — and offer to review the results afterwards.
Input: "I saved my swipe decisions yesterday — apply them."
Output: !bash ~/.claude/skills/skills-janitor/scripts/swipe.sh --apply ~/.skills-janitor/swipe-<timestamp>.json (list the files in ~/.skills-janitor/ to find the right one).
{baseDir}/../../scripts/swipe.sh; deck builder: {baseDir}/../../scripts/swipe-build-deck.sh~/.skills-janitor/log.jsonl/janitor-report — same data as a non-interactive list/janitor-value — the token + usage data underneath the swipe deck/janitor-fix --prune — automated broken-symlink cleanup, no interactive reviewCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.
Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.
Replace with description of the skill and when Claude should use it.
Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies
This skill should be used when the user wants to "create a skill", "add a skill to plugin", "write a new skill", "improve skill description", "organize skill content", or needs guidance on skill structure, progressive disclosure, or skill development best practices for Claude Code plugins.
Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.
Use when creating new skills, editing existing skills, or verifying skills work before deployment
Take khendzel/janitor-swipe from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.