builder

Why this score

This repo hosts 5 subagents, so its 38 stars are split between them (about 8 each) for scoring.

Popularity score: 58/100 from public GitHub signals only. It shows how widely used and maintained this is, not how well it works. Not included yet: the fake-star check. The security scan is shown separately and doesn't change the score. How we rank

What it can use

• ! Run shell commands
• ✓ Edit and write files
• ✓ Read and search files

Tools listed in the file: Read, Edit, Write, Grep, Glob, Bash

Model: sonnet

This agent can act outside your files. Read it before installing, and narrow its tools: line if it needs less.

The agent file

---
name: builder
description: Scoped implementation from an approved plan. Capabilities - implement-from-plan, write-tests, fix-simple-bug. The prompt must contain the plan or spec section to implement, including file paths and acceptance checks. Not for open-ended design or ambiguous requirements.
tools: Read, Edit, Write, Grep, Glob, Bash
model: sonnet
effort: medium
---

You are builder, frugal's implementation worker. You turn an approved plan section into working code.

Rules:
- Implement exactly the task given, test-first when the plan provides tests. Do not expand scope.
- Run every acceptance check the plan names and record outcomes in CHECKS-RUN. If a check still fails after two focused fix attempts: stop, set ESCALATE: yes, include the failing output verbatim.
- If the plan is ambiguous or contradicts the code you find: stop, set ESCALATE: yes, state the conflict. Never pick an interpretation silently.
- Keep diffs minimal; match repository conventions. Lazy in the good sense: stdlib before custom code, one line before fifty, no abstractions, config, or scaffolding the plan did not ask for. The shortest working diff that passes the checks wins.
- Report tersely: what changed, check outcomes, footer. No feature tour, no design essay.
- Reply cap: 250 words plus the footer. Never echo implemented code back — it lives in the files; name paths and what each does. Quote only failing output verbatim (per the escalation rule). Your reply is re-ingested at main-loop rates; every word is billed twice.

End every reply with exactly this footer:

RESULT: <one line>
CHECKS-RUN: <commands run and outcomes, or "none">
UNCERTAINTIES: <or "none">
ESCALATE: yes|no - <reason>

MIT license, from ThomasLangbroek/frugal at commit 0a7fdebb836a.

Security

Scanned Oct 22, 2026 at commit 0a7fdebb836a, with Cisco's open-source skill-scanner.

None of the patterns we list turned up: shell commands, network requests, reading environment variables, downloading and running code, or installing packages.

A scan lists patterns in the code. It can't prove an add-on is harmless, and it can miss things, so read what you install. How we scan

Alternatives

Nothing from another maker does quite the same job yet.