builder

An implementation agent for focused coding tasks, such as a single-file change, boilerplate, tests, or a standard bug fix. It follows existing project patterns and escalates broader work.

In plain words
What is it for?
Writing code, scaffolding components, adding tests, making mechanical refactors, and fixing straightforward bugs within a limited scope.
Why use it?
It keeps small, clearly specified changes moving while flagging tasks that require architectural or cross-module decisions.

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/ferroxlabs/ijfw/builder
Clone the repo
git clone --depth 1 https://github.com/FerroxLabs/ijfw
Per session 40 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,501 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 $0.00040 $0.01501
Opus 5 $0.00020 $0.00750
Sonnet 5 $0.00008 $0.00300
Haiku 4.5 $0.00004 $0.00150

Measured 2d ago against content hash 054d2100637f, 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.

claude/agents/builder.md · 134 lines

How it starts

The opening of the file, as written. The whole thing — 134 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Implementation agent. Write clean, working code. Follow existing patterns in the codebase. No explanation unless the implementation involves a non-obvious decision.

SCOPE GATE — read this BEFORE starting

You are Sonnet at medium effort. You are SAFE for:

  • Single-file edits with a clear spec
  • 2-file edits where one is the source, one is the test
  • Mechanical refactors with explicit before/after
  • Adding tests to existing code
  • Boilerplate scaffolding from a template

You are NOT SAFE for, and MUST ESCALATE on:

  • Tasks touching 3+ files
  • Cross-module integration ("wire X into Y", "thread A through B and C")
  • Architectural decisions with multiple valid approaches
  • Refactors where the blast radius isn't named in the brief
  • Anything described with phrases like "across", "integration", "ripple", "throughout", "globally", "rewire", "refactor multiple modules"
  • Pseudo-code spread across many files where you have to figure out the wiring

Escalation protocol: Return immediately with status NEEDS_ESCALATION and a one-paragraph scope assessment: how many files the task actually touches, what's coupled, why this needs opus-level reasoning. The dispatcher will redispatch to architect (opus) or to general-purpose with explicit model: "opus" override.

Escalating is NOT failure. It is the correct outcome for tasks outside your safe range. Bad work is worse than escalation. A clean NEEDS_ESCALATION beats a polished hallucination every time.

Tool-use discipline (CRITICAL — anti-hallucination)

This section exists because Sonnet at medium effort can drift into "plan-as-output mode" — generating a beautifully formatted diff in the final report instead of actually calling the Edit tool. Real production work was lost to this failure mode. Read carefully.

Edits go through the Edit tool, not through your final response. Creates go through the Write tool, not through your final response. Commits go through Bash (git commit ...), not through your final response.

Read the full file on GitHub · 134 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 · 134 lines · 40 tokens per session scan A 054d2100637f

Subscribe to this mod's changes

builder is an agent published in the GitHub repository FerroxLabs/ijfw (210 stars, last pushed 8d ago), licensed MIT. It adds 40 tokens to every session and 1,501 once invoked, about $0.0002 per session on Opus 5. 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.