egoxx1/unified-master-harness
One policy, six skills and real safety hooks for Claude Code, Codex and Antigravity — measured, not assumed.
field: Developer tools workflow: Data / ETLFrontend & designCI/CD & deployPlanning & specsPR / code review
stack: Python
ARCHETYPE
YOLO Cowboy
Permissions? Never heard of 'em. Ships at the speed of `--dangerously-skip-permissions`.
GUARDRAILS
4/5
Blocks destructive commands · Protects secrets · Pre-tool screening hook · Sandbox or ask-first rules · details
Quality 24/100 Guardrails v2
Copy this rig
# review before running: this installs third-party code
$ npx degit egoxx1/unified-master-harness/.claude ./rig-unified-master-harness # inspect, then merge into .claude/ 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 egoxx1/unified-master-harness/.agents/skills/dev-debug .claude/skills/dev-debug $ npx degit egoxx1/unified-master-harness/.agents/skills/dev-plan .claude/skills/dev-plan $ npx degit egoxx1/unified-master-harness/.agents/skills/dev-review .claude/skills/dev-review $ npx degit egoxx1/unified-master-harness/.agents/skills/dev-ui-verify .claude/skills/dev-ui-verify $ npx degit egoxx1/unified-master-harness/.agents/skills/dev-verify .claude/skills/dev-verify $ npx degit egoxx1/unified-master-harness/.agents/skills/project-context .claude/skills/project-context
$ curl -fsSL --create-dirs -o .claude/agents/dev-reviewer.md https://raw.githubusercontent.com/egoxx1/unified-master-harness/main/.claude/agents/dev-reviewer.md Merge into .claude/settings.json (project) or ~/.claude/settings.json (user). Hook commands reference scripts in the source repo — copy those too.
{
"permissions": {
"deny": [
"Read(.env)",
"Read(.env.local)",
"Read(.env.*.local)",
"Read(.env.development)",
"Read(.env.dev)",
"Read(.env.production)",
"Read(.env.prod)",
"Read(.env.staging)",
"Read(.env.test)",
"Read(.envrc)",
"Read(*.pem)",
"Read(*.key)",
"Read(*.p12)",
"Read(*.pfx)",
"Read(id_rsa)",
"Read(id_ed25519)",
"Read(id_ecdsa)",
"Read(.netrc)",
"Read(.pgpass)",
"Read(.git-credentials)"
],
"ask": [
"Bash(git push *)",
"Bash(npm publish *)",
"Bash(pnpm publish *)",
"Bash(yarn publish *)",
"Bash(cargo publish *)",
"Bash(twine upload *)",
"Bash(docker push *)",
"Bash(gh pr merge *)",
"Bash(gh release *)",
"Bash(terraform apply *)",
"Bash(terraform destroy *)",
"Bash(kubectl apply *)",
"Bash(kubectl delete *)",
"Edit(/AGENTS.md)",
"Edit(/CLAUDE.md)",
"Edit(/agent-core/hooks/**)",
"Edit(/agent-core/scripts/**)",
"Edit(/agent-core/evals/**)",
"Edit(/agent-core/growth/policy.json)",
"Edit(/.codex/**)",
"Edit(/.agents/**)"
]
},
"hooks": {
"PreToolUse": [
{
"matcher": "Bash|PowerShell|Read|Grep|mcp__.*",
"hooks": [
{
"type": "command",
"command": "python3 -B \"${CLAUDE_PROJECT_DIR}/agent-core/hooks/claude_pretool.py\""
}
]
}
]
}
} Skills (6)
Subagents (1)
| dev-reviewer model: inherit | Independent fresh-context reviewer for a completed change that is high-risk (auth/security, payments, destructive data migration, public API/protocol, concurrency/shared state, production deployment l |
Hooks (5)
| event | matcher | runs |
|---|---|---|
| PreToolUse | Bash|PowerShell|Read|Grep|mcp__.* | python3 -B "${CLAUDE_PROJECT_DIR}/agent-core/hooks/claude_pretool.py" |
| PostToolUse | ^(Edit|Write|MultiEdit|NotebookEdit|Bash|Skill)$|^mcp__.*([Pp]laywright|[Bb]rowser) | python3 -B "${CLAUDE_PROJECT_DIR}/agent-core/hooks/claude_session.py" |
| Stop | * | python3 -B "${CLAUDE_PROJECT_DIR}/agent-core/hooks/claude_session.py" |
| PreCompact | * | python3 -B "${CLAUDE_PROJECT_DIR}/agent-core/hooks/claude_session.py" |
| SessionStart | * | python3 -B "${CLAUDE_PROJECT_DIR}/agent-core/hooks/claude_session.py" |
Slash commands (3)
/harness-check/retro/smoke
Permissions
deny (20)
Read(.env)
Read(.env.local)
Read(.env.*.local)
Read(.env.development)
Read(.env.dev)
Read(.env.production)
Read(.env.prod)
Read(.env.staging)
Read(.env.test)
Read(.envrc)
Read(*.pem)
Read(*.key)
Read(*.p12)
Read(*.pfx)
Read(id_rsa)
Read(id_ed25519)
Read(id_ecdsa)
Read(.netrc)
Read(.pgpass)
Read(.git-credentials)
ask (21)
Bash(git push *)
Bash(npm publish *)
Bash(pnpm publish *)
Bash(yarn publish *)
Bash(cargo publish *)
Bash(twine upload *)
Bash(docker push *)
Bash(gh pr merge *)
Bash(gh release *)
Bash(terraform apply *)
Bash(terraform destroy *)
Bash(kubectl apply *)
Bash(kubectl delete *)
Edit(/AGENTS.md)
Edit(/CLAUDE.md)
Edit(/agent-core/hooks/**)
Edit(/agent-core/scripts/**)
Edit(/agent-core/evals/**)
Edit(/agent-core/growth/policy.json)
Edit(/.codex/**)
Edit(/.agents/**)
allow (5)
Bash
Edit
Write
WebFetch
WebSearch
Similar rigs
nchandar/dev-kit
Skills and small bash tools that make Claude Code, Codex and OpenCode work the same way: grill, plan, build test-first, prove it, learn from it.
Pragmatist 1.3k tok · · Q3
Nymphium/settings
dotfiles, and other
Pragmatist 9.6k tok · · Q3
reyxdz/going-dark
One strict way to run a software project with any AI coding agent: issues as specs, handoff prompts, audits that catch every problem down to a typo, and fixes for what they find.
Pragmatist 1.0k tok · · Q0
95gabor/agentic-workflow
🔨 Agentic workflow template
Skill Collector 4.5k tok · · Q0