~ / rigs / nnguyenquangg / harness-engineering-kit

nnguyenquangg/harness-engineering-kit

The public part of a harness engineering setup for Claude Code: spec-first workflow, three parallel AI reviewers, UI walkthroughs verified through the real API, draft PRs with evidence, and a layered harness template.

↗ GitHub ★ 23 MIT updated 14d ago personal setup Claude Code
share on X
ARCHETYPE
Skill Collector
Ten-plus skills loaded on demand. A procedural-knowledge library.
CONTEXT TAX · EVERY TURN
~1.3k tokens
Featherweight · median rig: 2.3k · breakdown
GUARDRAILS
0/5
No committed guardrails · details
Quality 0/100 Guardrails v2

Copy this rig

# review before running: this installs third-party code
$ npx degit nnguyenquangg/harness-engineering-kit/agents ./rig-harness-engineering-kit/agents
$ npx degit nnguyenquangg/harness-engineering-kit/commands ./rig-harness-engineering-kit/commands
$ npx degit nnguyenquangg/harness-engineering-kit/skills ./rig-harness-engineering-kit/skills

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 nnguyenquangg/harness-engineering-kit/skills/autopilot .claude/skills/autopilot
$ npx degit nnguyenquangg/harness-engineering-kit/skills/diagnose .claude/skills/diagnose
$ npx degit nnguyenquangg/harness-engineering-kit/skills/fix .claude/skills/fix
$ npx degit nnguyenquangg/harness-engineering-kit/skills/implement-specs .claude/skills/implement-specs
$ npx degit nnguyenquangg/harness-engineering-kit/skills/make-plan .claude/skills/make-plan
$ npx degit nnguyenquangg/harness-engineering-kit/skills/open-pr .claude/skills/open-pr
$ npx degit nnguyenquangg/harness-engineering-kit/skills/research .claude/skills/research
$ npx degit nnguyenquangg/harness-engineering-kit/skills/spec-loop .claude/skills/spec-loop
$ npx degit nnguyenquangg/harness-engineering-kit/skills/task-analyzer .claude/skills/task-analyzer
$ npx degit nnguyenquangg/harness-engineering-kit/skills/ui-walkthrough .claude/skills/ui-walkthrough
$ npx degit nnguyenquangg/harness-engineering-kit/skills/what-now .claude/skills/what-now
$ curl -fsSL --create-dirs -o .claude/agents/diff-reviewer.md https://raw.githubusercontent.com/nnguyenquangg/harness-engineering-kit/main/agents/diff-reviewer.md
$ curl -fsSL --create-dirs -o .claude/agents/edge-case-reviewer.md https://raw.githubusercontent.com/nnguyenquangg/harness-engineering-kit/main/agents/edge-case-reviewer.md
$ curl -fsSL --create-dirs -o .claude/agents/security-reviewer.md https://raw.githubusercontent.com/nnguyenquangg/harness-engineering-kit/main/agents/security-reviewer.md
$ curl -fsSL --create-dirs -o .claude/agents/walkthrough-runner.md https://raw.githubusercontent.com/nnguyenquangg/harness-engineering-kit/main/agents/walkthrough-runner.md

This rig commits no guardrails. Here is the community baseline instead — the deny/ask rules most often found across all 16.2k rigs:

{
  "permissions": {
    "deny": [
      "Read(./.env)",
      "Read(./.env.*)",
      "Read(**/.env)",
      "Read(~/.ssh/**)",
      "Bash(git push --force:*)",
      "Read(**/*.pem)",
      "Read(.env)",
      "Read(~/.aws/**)",
      "Read(**/*.key)",
      "Read(**/.env.*)",
      "Bash(git push --force*)",
      "Bash(rm -rf:*)",
      "Bash(rm -rf *)",
      "Bash(sudo *)",
      "Bash(sudo:*)",
      "Bash(rm -rf /)",
      "Bash(git push -f:*)",
      "Read(.env.*)",
      "Bash(git reset --hard*)",
      "Bash(rm -rf /*)"
    ],
    "ask": [
      "Bash(git push:*)",
      "Bash(git push *)",
      "Bash(gh pr merge *)",
      "Bash(git commit:*)",
      "Bash(rm *)",
      "Bash(gh pr merge:*)",
      "Bash(gh pr create *)",
      "Bash(git reset --hard:*)",
      "Bash(npm publish:*)",
      "Bash(git rebase *)"
    ]
  }
}

Skills (11)

Subagents (4)

diff-reviewerRead-only review of a diff for correctness against the task and the project's standards - missing or wrong requirements, scope creep, leftovers the change made redundant (constant flags, dead branches
edge-case-reviewerRead-only hunt for edge cases a diff mishandles - null/empty/missing values, boundaries and limits, concurrency and retries, partial failure, ordering and idempotency, config gaps, mixed old/new deplo
security-reviewerRead-only security review of a diff - authn/authz gaps, injection, secrets and PII exposure, unsafe input handling, SSRF, brute-force and enumeration, crypto and session misuse, dependency risk. One o
walkthrough-runnerRuns a prepared UI walkthrough (a throwaway Playwright walk.mjs) against a LOCAL running app, reads every screenshot and writes review.md with one OK/FINDING line per shot, then reports. Launched in p

Slash commands (8)

/autopilot/fix/implement-specs/make-plan/research/spec-loop/task-analyzer/what-now

Similar rigs

copied ✓