ToddGBenson/keel
A governed, agent-driven SDLC you can fork. NIST 800-53 Rev.5 + CIS Supply Chain + AI RMF. Gates with evidence, not assertions.
ARCHETYPE
Fort Knox
Deny lists, pre-tool hooks, sandboxing. Nothing touches prod without a signature.
GUARDRAILS
5/5
Blocks destructive commands · Protects secrets · Pre-tool screening hook · No YOLO mode · Sandbox or ask-first rules · details
Quality 54/100 Guardrails v2
Copy this rig
# review before running: this installs third-party code
$ npx degit ToddGBenson/keel/.claude ./rig-keel # 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 ToddGBenson/keel/.claude/skills/ai-impact-assessment .claude/skills/ai-impact-assessment $ npx degit ToddGBenson/keel/.claude/skills/artifact-intake .claude/skills/artifact-intake $ npx degit ToddGBenson/keel/.claude/skills/backlog-prioritization .claude/skills/backlog-prioritization $ npx degit ToddGBenson/keel/.claude/skills/blameless-postmortem .claude/skills/blameless-postmortem $ npx degit ToddGBenson/keel/.claude/skills/control-assessment .claude/skills/control-assessment $ npx degit ToddGBenson/keel/.claude/skills/control-verification .claude/skills/control-verification $ npx degit ToddGBenson/keel/.claude/skills/dependency-vetting .claude/skills/dependency-vetting $ npx degit ToddGBenson/keel/.claude/skills/eval-design .claude/skills/eval-design $ npx degit ToddGBenson/keel/.claude/skills/evidence-writing .claude/skills/evidence-writing $ npx degit ToddGBenson/keel/.claude/skills/exploratory-testing .claude/skills/exploratory-testing $ npx degit ToddGBenson/keel/.claude/skills/red-teaming .claude/skills/red-teaming $ npx degit ToddGBenson/keel/.claude/skills/retro-facilitation .claude/skills/retro-facilitation $ npx degit ToddGBenson/keel/.claude/skills/scanner-triage .claude/skills/scanner-triage $ npx degit ToddGBenson/keel/.claude/skills/secure-code-review .claude/skills/secure-code-review $ npx degit ToddGBenson/keel/.claude/skills/story-splitting .claude/skills/story-splitting $ npx degit ToddGBenson/keel/.claude/skills/test-strategy .claude/skills/test-strategy $ npx degit ToddGBenson/keel/.claude/skills/threat-modeling .claude/skills/threat-modeling $ npx degit ToddGBenson/keel/.claude/skills/writing-acceptance-criteria .claude/skills/writing-acceptance-criteria
$ curl -fsSL --create-dirs -o .claude/agents/ai-developer.md https://raw.githubusercontent.com/ToddGBenson/keel/main/.claude/agents/ai-developer.md $ curl -fsSL --create-dirs -o .claude/agents/ai-risk-officer.md https://raw.githubusercontent.com/ToddGBenson/keel/main/.claude/agents/ai-risk-officer.md $ curl -fsSL --create-dirs -o .claude/agents/architect.md https://raw.githubusercontent.com/ToddGBenson/keel/main/.claude/agents/architect.md $ curl -fsSL --create-dirs -o .claude/agents/backend-developer.md https://raw.githubusercontent.com/ToddGBenson/keel/main/.claude/agents/backend-developer.md $ curl -fsSL --create-dirs -o .claude/agents/compliance-officer.md https://raw.githubusercontent.com/ToddGBenson/keel/main/.claude/agents/compliance-officer.md $ curl -fsSL --create-dirs -o .claude/agents/delivery-lead.md https://raw.githubusercontent.com/ToddGBenson/keel/main/.claude/agents/delivery-lead.md $ curl -fsSL --create-dirs -o .claude/agents/developer.md https://raw.githubusercontent.com/ToddGBenson/keel/main/.claude/agents/developer.md $ curl -fsSL --create-dirs -o .claude/agents/frontend-developer.md https://raw.githubusercontent.com/ToddGBenson/keel/main/.claude/agents/frontend-developer.md $ curl -fsSL --create-dirs -o .claude/agents/infrastructure-developer.md https://raw.githubusercontent.com/ToddGBenson/keel/main/.claude/agents/infrastructure-developer.md $ curl -fsSL --create-dirs -o .claude/agents/product-owner.md https://raw.githubusercontent.com/ToddGBenson/keel/main/.claude/agents/product-owner.md $ curl -fsSL --create-dirs -o .claude/agents/qa-engineer.md https://raw.githubusercontent.com/ToddGBenson/keel/main/.claude/agents/qa-engineer.md $ curl -fsSL --create-dirs -o .claude/agents/release-manager.md https://raw.githubusercontent.com/ToddGBenson/keel/main/.claude/agents/release-manager.md $ curl -fsSL --create-dirs -o .claude/agents/security-engineer.md https://raw.githubusercontent.com/ToddGBenson/keel/main/.claude/agents/security-engineer.md $ curl -fsSL --create-dirs -o .claude/agents/tech-writer.md https://raw.githubusercontent.com/ToddGBenson/keel/main/.claude/agents/tech-writer.md $ curl -fsSL --create-dirs -o .claude/agents/ux.md https://raw.githubusercontent.com/ToddGBenson/keel/main/.claude/agents/ux.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.*)",
"Read(./**/*.pem)",
"Read(./**/*.key)",
"Read(./**/*.p12)",
"Read(./**/*.pfx)",
"Read(./**/id_rsa*)",
"Read(./**/id_ed25519*)",
"Read(./**/credentials)",
"Read(./**/.aws/**)",
"Read(./**/.ssh/**)",
"Read(./**/*secret*.json)",
"Read(./**/*credentials*.json)",
"Read(./secrets/**)",
"Edit(./evidence/**/*.sarif)",
"Edit(./evidence/**/*.json)",
"Edit(./evidence/**/*.xml)",
"Write(./evidence/**/*.sarif)",
"Write(./evidence/**/*.json)",
"Write(./evidence/**/*.xml)",
"Bash(git push --force*)",
"Bash(git push -f*)",
"Bash(git reset --hard*)",
"Bash(git rebase*)",
"Bash(git filter-branch*)",
"Bash(git commit --no-verify*)",
"Bash(git commit -n *)",
"Bash(git push --no-verify*)",
"Bash(gh pr merge*)",
"Bash(gh pr review --approve*)",
"Bash(gh release create*)",
"Bash(gh workflow run*release*)",
"Bash(curl:*)",
"Bash(wget:*)"
],
"ask": [
"Bash(git push:*)",
"Bash(gh pr create:*)",
"Bash(gh issue close:*)",
"Bash(npm install:*)",
"Bash(npm i:*)",
"Bash(pip install:*)",
"Bash(go get:*)",
"Bash(cargo add:*)",
"Edit(./.github/workflows/**)",
"Write(./.github/workflows/**)",
"Edit(./.claude/agents/**)",
"Edit(./.claude/settings.json)",
"Edit(./process/gates/**)",
"Edit(./docs/compliance/**)"
]
},
"hooks": {
"PreToolUse": [
{
"matcher": "Bash",
"hooks": [
{
"type": "command",
"command": "bash \"$CLAUDE_PROJECT_DIR/.claude/hooks/guard-bash.sh\""
}
]
},
{
"matcher": "Write|Edit",
"hooks": [
{
"type": "command",
"command": "bash \"$CLAUDE_PROJECT_DIR/.claude/hooks/guard-write.sh\""
}
]
}
]
}
} Skills (18)
ai-impact-assessmentartifact-intakebacklog-prioritizationblameless-postmortemcontrol-assessmentcontrol-verificationdependency-vettingeval-designevidence-writingexploratory-testingred-teamingretro-facilitationscanner-triagesecure-code-reviewstory-splittingtest-strategythreat-modelingwriting-acceptance-criteria
Subagents (15)
| ai-developer model: opus | Implements AI-bearing features — retrieval pipelines, agent workflows, embeddings, model integrations — with untrusted content treated as data and behaviour covered by evals rather than assertions. Us |
| ai-risk-officer model: opus | Owns AI-specific risk in both directions — AI features the team ships, and the AI agents the team builds with. Runs AI impact assessments at G2, eval suites and red-team review at G4, monitors drift, |
| architect model: opus | Owns how it fits together — technical approach, NFRs, system decomposition, and allocation of security controls to specific components. Use for G2 design review, ADRs, and interface contracts. Writes |
| backend-developer model: opus | Implements services, APIs, business logic and data access test-first, with authorization enforced on the object rather than the route. Use for server-side stories. Never reviews, verifies, or releases |
| compliance-officer model: opus | Owns the control map and the truth of it. Verifies evidence exists and is intact for every claimed control, runs periodic assessments, maintains the POA&M and SSP. Assesses; never implements. Reports |
| delivery-lead model: opus | Orchestrator and Scrum Master. Owns flow and gate transitions — no work advances between gates without its approval. Removes impediments, protects WIP limits, runs retrospectives, and surfaces where t |
| developer model: opus | Implements G1-ready stories test-first, on a branch, with control evidence and an AI-authorship declaration. Use for all feature, bug, and chore implementation. Never reviews, verifies, or releases it |
| frontend-developer model: opus | Implements user-facing surfaces test-first — components, state, routing, client-side accessibility — against a UX specification it did not write. Use for any story whose change is visible to a person. |
| infrastructure-developer model: opus | Implements infrastructure as code — containers, manifests, Terraform, pipeline configuration — with least privilege and a rehearsed rollback. Use for deployment topology and CI/CD changes. Never deplo |
| product-owner model: opus | Owns what and why. Converts raw ideas into problem statements, epics into INVEST stories, and vague wishes into testable acceptance criteria. Use for intake triage, refinement, backlog prioritization, |
| qa-engineer model: opus | Owns evidence that it works. Defines test strategy per story, writes tests tracing to acceptance criteria, runs exploratory charters, maintains the AC-to-test traceability matrix, and co-approves G4. |
| release-manager model: opus | Owns change control. Assembles the change record, verifies every included item cleared G4, confirms rollback is rehearsed, and presents the release for human authorization at G5. Prepares; a human aut |
| security-engineer model: opus | Owns threat identification and control verification. Runs STRIDE at G2, verifies allocated controls are actually implemented at G4, triages scanner output into real findings, maintains POA&M entries. |
| tech-writer model: opus | Owns comprehensibility. User docs, runbooks, API references, onboarding material, and changelogs. Documents what exists and files a defect when behavior and intent disagree. Never invents behavior. |
| ux model: opus | Owns whether the software can actually be used. Personas, user journeys, wireframes, interaction and content design, and accessibility conformance. Co-approves G2 for usability and accessibility. Desi |
Hooks (3)
| event | matcher | runs |
|---|---|---|
| PreToolUse | Bash | bash "$CLAUDE_PROJECT_DIR/.claude/hooks/guard-bash.sh" |
| PreToolUse | Write|Edit | bash "$CLAUDE_PROJECT_DIR/.claude/hooks/guard-write.sh" |
| SessionStart | * | bash "$CLAUDE_PROJECT_DIR/.claude/hooks/session-brief.sh" |
Slash commands (17)
/ai-gate/design/idea/implement/learn/plan/promote/qa-gate/ready/refine/release/retro/review/security-gate/self-review/spec/status
Permissions
deny (34)
Read(./.env)
Read(./.env.*)
Read(./**/*.pem)
Read(./**/*.key)
Read(./**/*.p12)
Read(./**/*.pfx)
Read(./**/id_rsa*)
Read(./**/id_ed25519*)
Read(./**/credentials)
Read(./**/.aws/**)
Read(./**/.ssh/**)
Read(./**/*secret*.json)
Read(./**/*credentials*.json)
Read(./secrets/**)
Edit(./evidence/**/*.sarif)
Edit(./evidence/**/*.json)
Edit(./evidence/**/*.xml)
Write(./evidence/**/*.sarif)
Write(./evidence/**/*.json)
Write(./evidence/**/*.xml)
Bash(git push --force*)
Bash(git push -f*)
Bash(git reset --hard*)
Bash(git rebase*)
Bash(git filter-branch*)
Bash(git commit --no-verify*)
Bash(git commit -n *)
Bash(git push --no-verify*)
Bash(gh pr merge*)
Bash(gh pr review --approve*)
Bash(gh release create*)
Bash(gh workflow run*release*)
Bash(curl:*)
Bash(wget:*)
ask (14)
Bash(git push:*)
Bash(gh pr create:*)
Bash(gh issue close:*)
Bash(npm install:*)
Bash(npm i:*)
Bash(pip install:*)
Bash(go get:*)
Bash(cargo add:*)
Edit(./.github/workflows/**)
Write(./.github/workflows/**)
Edit(./.claude/agents/**)
Edit(./.claude/settings.json)
Edit(./process/gates/**)
Edit(./docs/compliance/**)
allow (15)
Bash(git status)
Bash(git diff:*)
Bash(git log:*)
Bash(git branch:*)
Bash(git switch:*)
Bash(git checkout -b:*)
Bash(git add:*)
Bash(gh issue view:*)
Bash(gh issue list:*)
Bash(gh pr view:*)
Bash(gh pr list:*)
Bash(gh pr diff:*)
Bash(gh run view:*)
Bash(gh run list:*)
Bash(gh api:*)
Similar rigs
yeyo11/agentry
Run Claude Code as a service: REST API, web UI and multi-agent orchestration in one container
Pragmatist 5.2k tok · · Q18
addit-digital/addit-harness
Config harness for Claude Code — subagents, language conventions, an engineering-loop process, ADRs, and a one-line auto-detecting install
Orchestrator 3.6k tok · · Q15
erayendes/kortext
The project brain for AI-driven development: a brief becomes approved analysis documents, then your own agent CLI (Claude, Codex etc.) writes the code. Local-first, no API key of its own.
Orchestrator 1.2k tok · · Q5
bsrikumar855-dot/ClaudeCODE-Autonomous-skills
An autonomous AI development skill for Claude Code. Turn project goals into planned, implemented, tested, and reviewed software using specialist AI subagents.
Orchestrator 789 tok · · Q0