
Swatkinson Toolkit
- Updated July 7, 2026
- swatkinson/swatkinson-cc-toolkit
swatkinson-toolkit bundles handle-it, which turns a Linear issue into a review-ready PR; claudecodile-review, an iterate-until-5/5 review-and-fix loop; issue-watcher, a live Linear status board; and skill-evaluate, retrospective skill self-evaluation. It targets PR review and issue-driven delivery. It is a code-review and delivery toolkit with supporting subagents.
Key points
- Linear issue to review-ready PR
- Iterate-until-5/5 review loop
- Live Linear status board
- Skill self-evaluation
Swatkinson Toolkit by the numbers
- Data as of Jul 7, 2026 (Skillselion catalog sync)
/plugin marketplace add swatkinson/swatkinson-cc-toolkit/plugin install swatkinson-toolkit@swatkinson-cc-toolkitAdd your badge
Show developers this skill is listed on Skillselion. Paste this into your README.
| Last updated | July 7, 2026 |
|---|---|
| Repository | swatkinson/swatkinson-cc-toolkit ↗ |
What it does
Linear issue→review-ready PR orchestration, an iterate-until-5/5 review⇄fix loop, a live Linear status board, and retrospective skill self-evaluation.
README.md
Swatkinson's Claude Code Toolkit
Version 1.0.0
Install
/plugin marketplace add swatkinson/swatkinson-cc-toolkit
/plugin install swatkinson-toolkit
To update later:
/plugin marketplace update swatkinson-cc-toolkit
The /plugin menu shows when a newer version is available (driven by the version in the plugin manifest).
What's inside — swatkinson-toolkit
Skills
| Skill | What it does |
|---|---|
handle-it |
End-to-end orchestrator: tracked issue (or freeform feature) → plan → implement → draft PR (opened in-house) → review⇄fix loop to double-5/5 (owns the fixer; review comes from a local pass or the repo's swat-reviewer GitHub Action) → CI + preview → auto-test → your manual review → ready-for-review. Project-generic — reads all project specifics from .claude/handle-it/. |
swat-review |
🪰 One in-house review pass on a PR: an Opus reviewer posts P#-tagged inline comments with suggested fixes and posts/updates the single rating comment scoring three facets (Code Quality, Spec. Adherence, Risk and Complexity). It doesn't fix or loop — re-run it (a human, handle-it, or a swat-reviewer GitHub Action) toward the double-5/5 gate. Comment formats from .claude/handle-it/rules/. |
handle-it-project-setup |
Scans a repo (build scripts, CLAUDE.md/AGENTS.md, CI workflows, PR template, available tracker tools) and generates the .claude/handle-it/ config (config.md + rules/*.md) that makes handle-it and swat-review work on that project. Run this once per repo. |
skill-evaluate |
Retrospective self-evaluation of a skill run, scored against a 6-dimension rubric. |
Agents (spawned by the skills)
swat-reviewer, swat-fixer, handle-it-shipper, handle-it-investigator, handle-it-test-runner.
Prerequisites
These skills were built for my workflow and lean on tools/plugins beyond this one. Install/configure these or the affected skills won't fully work:
- A project config —
.claude/handle-it/—handle-itandswat-revieware project-generic and read every project specific from this directory:config.md(verify gate, worktree command, tracker, branch naming, CI/preview, hard-rule files, migration tooling) plusrules/*.md(the PR title/description, commit-message, manual-review handoff, and 🪰 rating/inline-comment templates). Run/handle-it-project-setuponce per repo to generate it. Without it, those skills will prompt you to create it. - An issue tracker (optional) —
handle-itsupports Linear (via the Linear MCP server), GitHub Issues (viagh), Jira, or no tracker at all (freeform mode). The config declares which. agentsystem-coreplugin — the default implement/conflict/CI-fix engine skills (agentsystem-core:ship,:resolve-conflict,:fix-pr-tests). The config's Engine skills section can point at alternatives. (Opening the draft PR is now in-housed byhandle-it— no external open-PR skill needed.)diagnoseskill —handle-it's investigator uses it for unclear bugs. Comes from Matt Pocock's skills toolkit. (Code review is now in-housed inswat-review— it no longer needs the externalcode-review/simplifyskills.)- GitHub CLI (
gh), authenticated — all PR operations.
skill-evaluate is the only fully standalone skill — no external prerequisites.
Notes on portability
- Project specifics are no longer baked into the skills.
handle-it/swat-reviewhold zero project facts; everything lives in the per-repo.claude/handle-it/directory —config.mdfor facts,rules/*.mdfor the PR/commit/review/handoff templates (generated by/handle-it-project-setup). The same engine runs on a Bun/TanStack repo, a .NET solution, or an AutoCAD plugin — only the config differs. The engine also self-corrects both the config and the rule files when something proves wrong at runtime. - Some skill bodies reference paths like
~/.claude/skills/...and~/.claude/agents/...for orientation. After install via this plugin the files actually live under the plugin's directory; those references are documentation only and don't affect execution.
Repo layout
.claude-plugin/marketplace.json ← marketplace manifest (this repo)
plugins/swatkinson-toolkit/
.claude-plugin/plugin.json ← plugin manifest
skills/<skill>/SKILL.md ← + REFERENCE.md per skill
agents/<agent>.md
scripts/sync-version.ps1 ← keep the version in sync across the 3 files
Releasing
The version appears in three places that must stay in sync — plugins/swatkinson-toolkit/.claude-plugin/plugin.json (the source of truth), .claude-plugin/marketplace.json, and the subtitle at the top of this README. scripts/sync-version.ps1 keeps them aligned:
pwsh scripts/sync-version.ps1 # check: reports drift, exits 1 if the three disagree
pwsh scripts/sync-version.ps1 -Set 0.2.0 # bump: writes the new version to all three
Run the check before tagging a release (or wire it into a pre-commit hook / CI step).