gardening
Skill104 starsChanged 3 days agoApache-2.0View on GitHub
Weekly codebase groundskeeping. Sweeps the codebase for one or more hygiene pillars — hygiene (Prettier/import-sort/knip dead code), comments (low-value/commented-out), dry (dedup/reuse), refactor (leaner code), consistency (converge divergent implementations onto the canonical one), react (UI primitives + React Compiler), tests (coverage + e2e + stale-test repair) — applies only gate-verified fixes, and opens a DRAFT PR per pillar plus a queue issue (bot-owned carry-over + human decisions). Use for the weekly gardening run or when invoked as /gardening [pillars].
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 51
Read .claude/skills/gardening/pillars/<pillar>.md for that pillar's spec block (PILLAR_ID, CATEGORY_IDS, CONVENTION_SKILLS, SIGNALS, …
- Very specific: line 54
Run engine steps E3–E9 for that pillar (worklist → MAP → REDUCE → gate → apply/validate/revert → draft PR → self-review → queue update).
- Distinctive: line 57
Carry the claimed-files set forward: after a pillar opens a PR, add the files it touched to the claimed-files set so a later pillar in the …
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Codebase Gardening
Weekly groundskeeping for the Tangle-UI codebase. Sweep the tree for hygiene issues, apply only fixes
a validation gate can prove safe, self-review each PR against the review skill's criteria, and open draft PRs for
human review. The goal is a codebase that stays free of AI slop without depending on per-PR hand polish.
Work is organized into pillars, each a bounded scan run one at a time so a single run is unlikely to exhaust the token budget or time out:
| Pillar | What it does | Convention skill(s) |
hygiene | Fix Prettier/import-sort violations; remove knip dead code (runs first) | project-conventions, typescript-standards |
comments | Remove/refine low-value & commented-out comments; keep the why | project-conventions |
dry | Consolidate duplicated logic; adopt an existing helper before extract | project-conventions, typescript-standards |
refactor | Simplify verbose code; integrate with existing abstractions | typescript-standards, project-conventions |
consistency | Converge divergent implementations of one concept onto the canonical form (per canon-registry.md or a measured supermajority) | react-patterns, project-conventions, typescript-standards |
react | Raw HTML → UI primitives + tone; flag React Compiler adoption gaps | ui-primitives, react-patterns |
tests | Add missing tests, author e2e specs, repair stale tests (runs last) | vitest-testing, e2e-testing |
Arguments
Positional pillar names select what runs; flags modify the run.
- Pillars — any subset of
hygiene,comments,dry,refactor,consistency,react,tests, space-separated (e.g./gardening comments dry).all(or no pillar given) runs all seven in priority order. Unknown names → list the valid pillars and stop. --dry-run— analyze and print each selected pillar's worklist + findings table; create no branches, edits, commits, or PRs.--max-files N— overridecaps.maxFilesPerPRfor this run.--promote-compiler— allow thereactpillar to enable ready dirs inreact-compiler.config.js(otherwise React Compiler findings are flag-only). Ignored by other pillars.
How to run
- Parse arguments into
SELECTED_PILLARS(default: all seven, in config priority order —hygiene,comments,dry,refactor,consistency,react,testsper.github/gardening-config.json) and the flag set. - Read the shared engine once:
.claude/skills/gardening/engine.md. It defines steps E0–E10, the state model, fingerprints, and schemas. - Run the engine's shared preamble once (E0 config, E1 fork protection, E2 ISO-week + suppression/in-flight/file-claim scan). If E1 fails the repo check, stop for the whole run.
- For each pillar in
SELECTED_PILLARS, in order:- Read
.claude/skills/gardening/pillars/<pillar>.mdfor that pillar's spec block (PILLAR_ID,CATEGORY_IDS,CONVENTION_SKILLS,SIGNALS,SPECIAL_HANDLING). If the doc is missing, printSKIPPED <pillar>: missing pillar docand move to the next pillar — never improvise the spec. - Run engine steps E3–E9 for that pillar (worklist → MAP → REDUCE → gate → apply/validate/revert
→ draft PR → self-review → queue update). Each pillar produces its own branch
automated-gardening/<pillar>/<week>and its own draft PR. - Carry the claimed-files set forward: after a pillar opens a PR, add the files it touched to the claimed-files set so a later pillar in the same run cannot edit them (prevents mid-run collisions, complementing E2's scan of already-open PRs).
- Respect
caps.maxPRsPerRunacross the whole invocation, not per pillar.
- Read
- Run E10 cleanup once at the end (remove
/tmp/gardening-<week>, return to the starting branch).
Under --dry-run, step 4 stops after E6 for each pillar (print the table, create nothing) and step 5
still cleans up the temp dir.
Notes
- Loading is lazy by design: only the engine plus the selected pillar docs are read, so a single-pillar run stays small.
- The convention skills are the source of truth for every rule — this skill and its pillar docs cite them, never restate them.
- Everything this skill opens is a draft PR; nothing auto-merges.
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- joshrotenberg/practical-skills★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository20 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- accessibilitySkill
- address-pr-commentsSkill
- analytics-trackingSkill
- audit-ticketsSkill
- docs-updateSkill
- e2e-testingSkill
- list-skillsSkill
- open-sourceSkill
- project-conventionsSkill
- react-patternsSkill
- reviewSkill
- tangle-domainSkill
- tanstack-querySkill
- tanstack-routerSkill
- typescript-standardsSkill
- ui-primitivesSkill
- validateSkill
- vitest-testingSkill
- playwright-testingSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.