builder

A scheduled local Claude Code routine that turns a fully specified issue into a reviewed pull request, with one human approval step.

In plain words
What is it for?
It helps pick up ready issues, prepare an implementation plan, wait for approval, build the change, open a pull request, and leave merging to a human.
Why use it?
It separates planning, implementation, and review, while labels preserve the work's state between runs.

Agent

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/vscarpenter/gsd-task-manager/builder
Clone the repo
git clone --depth 1 https://github.com/vscarpenter/gsd-task-manager
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,380 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00000 $0.01380
Opus 5 $0.00000 $0.00690
Sonnet 5 $0.00000 $0.00276
Haiku 4.5 $0.00000 $0.00138

Measured 2d ago against content hash fceea1a1689b, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

builder scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

docs/agents/builder.md · 78 lines

How it starts

The opening of the file, as written. The whole thing — 78 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Builder — operating spec

The builder is a local, scheduled Claude Code routine (cycle B of the delivery pipeline). It turns a fully-specified issue into a reviewed pull request, pausing once for human plan approval (Gate 1). It never merges. It is launched by scripts/builder-run.sh; its per-run instructions live in .claude/commands/build-next.md; this file is the durable operating spec both reference.

For all issue-tracker operations, follow docs/agents/issue-tracker.md (the gh conventions). For triage-label vocabulary, see docs/agents/triage-labels.md.

The loop (label state machine)

The builder is non-blocking: each run does one unit of work and exits. Labels carry state between runs, so a closed laptop never orphans anything.

needs-triage + risk:*            (a filed contract, cycle A)
      │  human triages a well-specified one
      ▼
ready-for-agent ──run 1 (plan)──▶ plan:pending ──human swaps──▶ plan:approved ──run 2 (build)──▶ [PR opened]
      │                               │                                            │
      │ risk:docs | risk:chore        │ human swaps + /revise <notes>              │ agent:building
      │ auto-approve (plan+build       ▼                                           │  (claim-lock, removed
      │  in one pass, logged)     plan:revise ──run (re-plan)──▶ plan:pending      │  when the PR is opened)
      │
      └── ambiguous / needs judgment / hard limit hit ──▶ ready-for-human + written reason

Process at most one issue per run (one planned OR one built), so a single run has a bounded blast radius.

Risk → plan depth

The risk:* label (from the contract) decides how deep the plan is and whether Gate 1 applies:

Risk Plan depth Gate 1
risk:docs One or two lines: what changes, how verified. Auto-approved — post the plan as a comment noting auto-approved (risk:docs), then build in the same run.
risk:chore Short: approach + files touched + how verified. Auto-approved — same, auto-approved (risk:chore).
risk:feature Full plan: approach, files to touch, test strategy, open questions. Follow the structure in docs/superpowers/plans/. Required — post plan, label plan:pending, stop.
risk:risky Full plan plus an explicit risk/rollback section and the specific safeguards. Required — post plan, label plan:pending, stop.

Read the full file on GitHub · 78 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 78 lines · 0 tokens per session scan A fceea1a1689b

Subscribe to this mod's changes

builder is an agent published in the GitHub repository vscarpenter/gsd-task-manager (22 stars, last pushed 3d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,380 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other agents, from other repositories

mcp-tool-developer

MCP tool developer agent for building S&P 500 MCP server tools.

zhensherlock/sp500-mcp-server · 14 tokens

ot-architect

Usar para decisiones de arquitectura de OpenTicket — límites del core framework-free (core/), diseño del AgentCommerceAdapter, nuevos rails, extracción del paquete OSS, o cualquier cambio que cruce core/ ↔ lib/ ↔ app/. Invocar ANTES de escribir código que toque core/adapter o ports.ts.

LucasLeguizamo/open-ticket · 69 tokens

ot-db

Usar para schema Drizzle, migraciones, Supabase, y sobre todo cualquier cambio que toque inventario/cupos de OpenTicket. Invocar antes de modificar db/schema.ts o db/store.ts, y para operar la DB (push, seed, advisors de Supabase).

LucasLeguizamo/open-ticket · 59 tokens

ot-devops

Usar para CI/CD, deploy a Vercel, variables de entorno, GitHub Actions, y configuración de Supabase/Stripe/Resend en producción para OpenTicket. Invocar para el primer deploy, para armar el workflow de CI, o cuando falle un build/deploy.

LucasLeguizamo/open-ticket · 63 tokens

ot-frontend

Usar para la web humana de OpenTicket — landing dev-native estética CLI, ticker en vivo, página pública de evento, dashboard organizador, página de orden. Invocar para cualquier cambio de UI/UX en app/.

LucasLeguizamo/open-ticket · 49 tokens

ot-payments

Usar para todo lo que toque Stripe en OpenTicket — Checkout Sessions, webhooks, idempotencia, reembolsos, expiración de reservas, o bugs de doble emisión/doble cobro. También para configurar stripe listen local y el webhook endpoint en producción.

LucasLeguizamo/open-ticket · 59 tokens