Greg-Klein/implementation-harness
Local UI and Claude Code plugin that drives a GitLab ticket to a merge request, with agents, tools and deliverables kept visible
ARCHETYPE
Orchestrator
A bench of specialised subagents. The main agent mostly delegates.
Copy this rig
# review before running: this installs third-party code $ npx degit Greg-Klein/implementation-harness/agents ./rig-implementation-harness/agents $ npx degit Greg-Klein/implementation-harness/commands ./rig-implementation-harness/commands $ npx degit Greg-Klein/implementation-harness/skills ./rig-implementation-harness/skills $ npx degit Greg-Klein/implementation-harness/hooks ./rig-implementation-harness/hooks
MCP servers are added to Claude Code at local scope; env vars are shown as YOUR_… placeholders — we never store values. Files are fetched with degit into a separate folder so you can review before merging.
$ npx degit Greg-Klein/implementation-harness/skills/clarify-spec .claude/skills/clarify-spec $ npx degit Greg-Klein/implementation-harness/skills/collect-evidence .claude/skills/collect-evidence $ npx degit Greg-Klein/implementation-harness/skills/document-change .claude/skills/document-change $ npx degit Greg-Klein/implementation-harness/skills/figma-review .claude/skills/figma-review $ npx degit Greg-Klein/implementation-harness/skills/gh-github-api .claude/skills/gh-github-api $ npx degit Greg-Klein/implementation-harness/skills/gitlab-tickets .claude/skills/gitlab-tickets $ npx degit Greg-Klein/implementation-harness/skills/glab-gitlab-api .claude/skills/glab-gitlab-api $ npx degit Greg-Klein/implementation-harness/skills/how .claude/skills/how $ npx degit Greg-Klein/implementation-harness/skills/review-change .claude/skills/review-change $ npx degit Greg-Klein/implementation-harness/skills/self-check .claude/skills/self-check $ npx degit Greg-Klein/implementation-harness/skills/unslop .claude/skills/unslop $ npx degit Greg-Klein/implementation-harness/skills/why .claude/skills/why
$ curl -fsSL --create-dirs -o .claude/agents/designer-reviewer-figma.md https://raw.githubusercontent.com/Greg-Klein/implementation-harness/main/agents/designer-reviewer-figma.md $ curl -fsSL --create-dirs -o .claude/agents/developer.md https://raw.githubusercontent.com/Greg-Klein/implementation-harness/main/agents/developer.md $ curl -fsSL --create-dirs -o .claude/agents/qa-reviewer.md https://raw.githubusercontent.com/Greg-Klein/implementation-harness/main/agents/qa-reviewer.md $ curl -fsSL --create-dirs -o .claude/agents/review-orchestrator.md https://raw.githubusercontent.com/Greg-Klein/implementation-harness/main/agents/review-orchestrator.md $ curl -fsSL --create-dirs -o .claude/agents/senior-reviewer.md https://raw.githubusercontent.com/Greg-Klein/implementation-harness/main/agents/senior-reviewer.md $ curl -fsSL --create-dirs -o .claude/agents/ticket-planner.md https://raw.githubusercontent.com/Greg-Klein/implementation-harness/main/agents/ticket-planner.md $ curl -fsSL --create-dirs -o .claude/agents/ticket-scheduler.md https://raw.githubusercontent.com/Greg-Klein/implementation-harness/main/agents/ticket-scheduler.md
This rig commits no guardrails. Here is the community baseline instead — the deny/ask rules most often found across all 7,204 rigs:
{
"permissions": {
"deny": [
"Read(./.env)",
"Read(**/.env)",
"Read(~/.ssh/**)",
"Bash(rm -rf *)",
"Read(**/*.pem)",
"Bash(rm -rf /)",
"Bash(git push --force:*)",
"Bash(sudo *)",
"Read(.env)",
"Bash(rm -rf /*)",
"Read(./.env.*)",
"Read(~/.aws/**)",
"Bash(git push --force*)",
"Bash(rm -rf:*)",
"Read(**/*.key)",
"Read(**/.env.*)",
"Bash(sudo:*)",
"Bash(git reset --hard*)",
"Bash(git reset --hard:*)",
"Read(.env.*)"
],
"ask": [
"Bash(git push:*)",
"Bash(git push *)",
"Bash(git commit:*)",
"Bash(rm *)",
"Bash(rm:*)",
"Bash(git rebase *)",
"Bash(wget *)",
"Bash(npm publish:*)",
"Bash(git commit *)",
"Bash(gh pr merge *)"
]
}
} Skills (12)
clarify-speccollect-evidencedocument-changefigma-reviewgh-github-apigitlab-ticketsglab-gitlab-apihowreview-changeself-checkunslopwhy
Subagents (7)
| designer-reviewer model: sonnet | Independently review a change visible in the UI against Figma, ticket mockups or already shipped screens, measure visual, state, accessibility and layout defects in the live interface and return evide |
| developer model: sonnet | Implement an assigned plan task or rework within its file scope, self-check the change and return versioned evidence for independent review. |
| qa-reviewer model: opus | Independently select counterexample and acceptance checks on the final implementation, execute them and report observed results without fixing product code. |
| review-orchestrator model: sonnet | Sequence independent code, design and QA reviews, route scoped corrections and preserve versioned evidence until the bounded review loop ends. Does not implement, commit or push. |
| senior-reviewer model: opus | Independently challenge a change, then correct justified defects within its authorized scope and leave a report for final independent QA. |
| ticket-planner model: opus | Turn a software request and current code into a scoped, executable plan with explicit assumptions, dependencies, risks and verification steps. |
| ticket-scheduler model: sonnet | Predict which files and areas each GitLab or GitHub ticket of a batch would touch in the current repository, and which tickets overlap or depend on each other, without changing the repository. |
Slash commands (6)
/implement/improve/judge-improvement/rebase/review/schedule
Similar rigs
danjdewhurst/dotclaude
Global Claude Code config, synced between machines by one script: CLAUDE.md, writing rules, vendored skills and the bash environment commands run in, symlinked into ~/.claude and wherever Codex and other agents look
Pragmatist 398 tok ·
devon-systems/hoenn
NixOS, nix-darwin, and Home Manager configurations, ft. niri+noctlaia dotfiles + homelab.
Skill Collector 1.2k tok ·
emmaneugene/claude
Claude Code config
YOLO Cowboy 1.6k tok ·
firemonster612/dot-agents
My AI coding setup / agent dotfiles — skills, workflow, and config
Skill Collector 9.7k tok ·