
Fakoli Crew
- 4 repo stars
- Updated July 13, 2026
- fakoli/fakoli-plugins
Deploy eight specialist software-engineering agents (architect, reviewer, researcher, QA, etc.) with TDD and verification gates.
About
fakoli-crew supplies eight specialist AI agents for software engineering, including a polyglot architect, code reviewer, researcher, plugin engineer, integration specialist, documentation writer, infrastructure keeper, and QA validator. It enforces TDD, systematic debugging, and verification gates. Use it as a ready-made agent team to build and check software.
- Eight specialist agents
- Polyglot architect + reviewer
- TDD enforcement
- Verification gates
Fakoli Crew by the numbers
- Data as of Jul 14, 2026 (Skillselion catalog sync)
/plugin marketplace add fakoli/fakoli-plugins/plugin install fakoli-crew@fakoli-pluginsAdd your badge
Show developers this skill is listed on Skillselion. Paste this into your README.
| repo stars | ★ 4 |
|---|---|
| Last updated | July 13, 2026 |
| Repository | fakoli/fakoli-plugins ↗ |
What it does
Deploy eight specialist software-engineering agents (architect, reviewer, researcher, QA, etc.) with TDD and verification gates.
README.md
fakoli-crew
Nine specialized AI agents — generalist engineering roles (polyglot architect, Staff Engineer code reviewer, security auditor, API researcher, plugin engineer, integration specialist, documentation writer, infrastructure keeper, QA validator) — that run independently or as coordinated multi-wave crews for complex development projects.
Looking for plugin-development critics? The five specialist critics (
agent-critic,skill-critic,hook-critic,mcp-critic,structure-critic) moved to a dedicated plugin in 2.3.0. Installfakoli-plugin-criticalongside this one to get them. Existing recipes that dispatched viafakoli-crew:<critic>must update the namespace tofakoli-plugin-critic:<critic>.
Architecture

Critic Gate

Installation
claude plugin install fakoli-crew
Available Agents
| Agent | Role | Trigger Phrases |
|---|---|---|
| guido | TypeScript architect | "design an interface", "make this more TypeScript", "create an interface" |
| critic | Code reviewer | "review this code", "find the bugs", "audit the imports" |
| scout | Researcher | "explore this codebase", "look up the API docs", "map the dependencies" |
| smith | Plugin engineer | "update the manifest", "add a command", "set up the hooks" |
| welder | Integration engineer | "wire up the new", "refactor to use", "connect these modules" |
| herald | Documentation writer | "improve the README", "make this appealing", "rewrite the description" |
| keeper | Infrastructure engineer | "update CLAUDE.md", "fix the CI", "sync the registry" |
| sentinel | QA engineer | "validate everything", "run all tests", "check if this is ready" |
| warden | Security auditor | "security review", "audit the permission surface", "check for leaked secrets" |
Using Agents Individually
Each agent is a focused expert. Invoke one directly when you know what you need:
/agent:guido Design a Provider interface for the new TTS backend.
/agent:critic Review src/client.py for correctness and TypeScript style.
/agent:scout Read the Stripe API docs and map the endpoints we need.
/agent:smith Add a `tts generate` command to the plugin manifest.
/agent:welder Wire the new ProviderProtocol into the existing TTSClient.
/agent:herald Rewrite the README — make it appealing to first-time visitors.
/agent:keeper Sync CLAUDE.md and marketplace.json after the restructure.
/agent:sentinel Validate everything before we tag v2.0.0.
Pre-Built Crews
Use the crew skill when a task spans multiple concerns:
Code Quality
guido + critic + sentinel — audit and improve an existing codebase.
critic finds issues (with severity ratings) →
guido rewrites with idiomatic TypeScript alternatives →
sentinel validates with a pass/fail scorecard
Plugin Development
smith + guido + sentinel + herald — ship a new plugin or major feature.
scout researches existing patterns →
smith builds the manifest, guido builds the code →
sentinel validates, herald writes the README
Research & Build
scout + guido + welder (critic as gate) — integrate an external API you haven't used before.
scout maps the API →
guido designs the wrapper interface →
── CRITIC GATE ──
welder wires it in →
── CRITIC GATE ──
Documentation Sprint
herald + keeper + sentinel — rewrite docs, update infrastructure, verify consistency.
herald rewrites READMEs and descriptions →
── CRITIC GATE ──
keeper updates CLAUDE.md, CI, contributor docs →
sentinel verifies all sources are in sync
Full Overhaul
All 9 agents in 5 waves — major version bump, structural refactor, or public launch prep.
The Wave Pattern
Crews execute in waves to manage dependencies. No two agents write the same file.
Wave 1 — Research: scout gathers information
Wave 2 — Build (parallel): guido + smith + herald create new artifacts
── CRITIC GATE ── critic reviews all modified files (non-negotiable)
Wave 3 — Integrate: welder wires everything together
── CRITIC GATE ── critic reviews the integration
Wave 4 — Final Verification: sentinel produces evidence-based scorecard
Wave 5 — Infrastructure: keeper syncs CLAUDE.md, CI, registry
Each agent writes a status file to the path the orchestrator provides in its dispatch
prompt. The default scratch root is .fakoli/runs/<run-id>/agent-<name>-status.md
(gitignored). Agents coordinate through these files without a shared conversation.
Commands
| Command | Description |
|---|---|
/crew |
List all 9 agents and suggested crew compositions |
Skills
| Skill | Description | File |
|---|---|---|
| Debugging | Systematic 4-phase root cause analysis — used automatically by critic, welder, and sentinel when diagnosing failures | skills/debugging/SKILL.md |
| Crew-ops | Multi-agent orchestration — wave patterns, file ownership, status protocol, and shared language style references for coordinated work across the crew | skills/crew-ops/SKILL.md |
Agent Design Principles
guido applies TypeScript naming conventions, strict typing, interface-first design, and generics. Reviews with severity levels: MUST / SHOULD / CONSIDER / NIT.
critic reads line by line, proposes alternative implementations, and never approves code that swallows exceptions, uses mutable defaults, or lacks type annotations.
scout explores before concluding. Reads actual source files and official docs rather than guessing. Reports what it found, not what it expected to find.
smith follows the plugin manifest schema precisely. Validates every field before writing. Cross-references CLAUDE.md to confirm nothing is missing.
welder reads ALL upstream agent files before modifying anything. Maintains backward compatibility through re-exports and facade patterns. Runs tests after every change.
herald writes for strangers, not existing users. Starts with a concrete value proposition. Never writes "A tool for X" — always says what X specifically does.
keeper keeps every source of truth in sync: CLAUDE.md, marketplace.json, registry, CI workflows, and contributor docs. Makes surgical edits, never wholesale rewrites.
sentinel produces a binary pass/fail scorecard for every check. Reports exact error output. Names the fix owner for every failure. Never modifies code itself.
Works Best With
fakoli-flow — the crew-aware orchestrator. Install both plugins for the full workflow:
claude plugin install fakoli-crew
claude plugin install fakoli-flow
fakoli-flow handles wave dispatch, critic gates, and status file coordination automatically. See docs/workflow-orchestration.md for the comparison between fakoli-flow and SuperPowers orchestration.
Requirements
- Claude Code with plugin support
- No external dependencies — agents use built-in Claude Code tools only
Author
Sekou Doumbouya — github.com/fakoli
License
MIT — see LICENSE