- Home
- Skills
- Agents & Orchestration
- Claude Code Multi-Agent Pipeline: Spec to Verified Build
Works with the AI Agents you already use
Claude Code Multi-Agent Pipeline: Spec to Verified Build
Five role-restricted sub-agents, from planner to verifier. Hand-offs pass only when a command's exit code says so, never an agent's word.
$29
Claude Code Multi-Agent Pipeline: Spec to Verified Build
Example session with this skill installed
Build an idempotent POST /transfers endpoint for our ledger with the agent-forge pipeline.
- Read your context and instructions
- Compiled the claude code multi-agent
Built: idempotent double-entry ledger with POST /transfers. Gate report:
| Stage | Verdict | Loops |
|---|---|---|
| Spec | 9 acceptance criteria approved | 0 |
| Architect | 3 decision records + contract | 0 |
| Tasks | TRACE: PASS (every criterion has a task) | 0 |
| Implement | 20 tests pass | 0 |
| Review | GATE: FAIL: lock timeout leaks a raw OperationalError (HTTP 500); failing test committed | 1 |
| Fix | bounded retry + typed 503 with Retry-After | 0 |
| Verify | VERIFY: FAIL twice (coverage 68% < 80%, 11 lint findings), then VERIFY: PASS: 45 tests, 97.9% coverage | 2 |
| Trace | TRACE: PASS, 9/9 criteria → tasks → tests | 0 |
The reviewer's red test and the fix are separate commits, so the history proves the review mattered. Nothing was waived.
(From the public agent-forge run, summarised.)
Connects securely to your tools. The creator never sees your data.
About this skill
Multi-agent builds fail when agents grade their own homework. This pipeline makes each stage check the one before it.
What you get
- Five Claude Code sub-agents with ready-to-use role files: planner, architect, implementer, adversarial reviewer and verifier (tools restricted per role).
- Hard gates run by the orchestrator: every requirement traced to a real test, review findings backed by failing tests, every check decided by exit code.
- Bounded fix loops that never lower the bar to get green.
- Templates for spec, tasks, review and verification, plus attack-surface and stack-command references.
Battle-tested on a public payments ledger: the verify gate halted the build twice and the reviewer caught a real concurrency bug the implementer's tests missed.
Why this beats one long "build this feature" prompt One agent writes, reviews and approves its own work. Here each stage is a separate sub-agent with restricted tools, and the orchestrator runs the gate commands itself, so a stage cannot pass by saying it passed.
Use cases
- Building a feature or service end to end with Claude Code sub-agents
- Spec-driven work that must not silently drop requirements
- Auditing how AI-written code was reviewed and verified
Known limitations
- Built for Claude Code sub-agents; the gates and templates are reusable elsewhere.
- Slower and costlier than a single prompt: it spends tokens on review and verification on purpose.
- Needs a project with a runnable test command.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 28 days ago
Requires