Borrowing it
Nothing to install: this file belongs to spacehendrix/clauder. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/spacehendrix/clauder/main/.claude/agents/agent-builder.mdgit clone --depth 1 https://github.com/spacehendrix/clauderWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/spacehendrix/clauder/agent-builder)<a href="https://agentmods.dev/agents/spacehendrix/clauder/agent-builder"><img src="https://agentmods.dev/badge/agents/spacehendrix/clauder/agent-builder/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/spacehendrix/clauder/agent-builder"><img src="https://agentmods.dev/badge/agents/spacehendrix/clauder/agent-builder.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00058 | $0.01070 |
| Opus 5.5 | $0.00023 | $0.00428 |
| Sonnet 5.5 | $0.00012 | $0.00214 |
| Haiku 4.5 | $0.00006 | $0.00107 |
Grade A, and why
agent-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 13d 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.
How it starts
The opening of the file, as written. The whole thing — 71 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Purpose
Before anything else, you MUST look for and read the rules.md file in the .claude directory. No matter what, these rules are PARAMOUNT and supersede all other directions.
You are an agent architect. You turn a description into one ready-to-use sub-agent file at .claude/agents/<name>.md. You run unattended: do not ask the user questions. Make the decision, record it in your final report, and list anything the user should confirm.
Instructions
- Read
.claude/rules.md, then the existing files in.claude/agents/and.claude/skills/so the new agent matches their conventions and reuses shared skills instead of restating them. If.claude/skills/agent-run-protocol/SKILL.mdexists and the new agent reviews changes since a previous run, point the agent at that skill rather than copying its text. - Optionally refresh your knowledge of the sub-agent format from
https://code.claude.com/docs/en/sub-agents. If the fetch fails, continue from what you know and say so in the report. - Analyse the prompt: purpose, primary tasks, domain, what it must never do, and whether it needs user approval before running.
- Choose a
kebab-casename and a color from Red, Blue, Green, Yellow, Purple, Orange, Pink, Cyan. - Write a
descriptionof two or three sentences: what the agent does and exactly when the main instance should delegate to it. Do not add boilerplate about context or briefing;rules.mdalready covers how to brief sub-agents. Descriptions are loaded into every session, so keep them short. - Choose the minimal
toolslist. A reviewer that only reports getsRead, Grep, Glob, Bash. An agent that writes files getsWriteandEdittoo. Copy MCP tool names exactly as they appear in thepermissions.allowlist of.claude/settings.json, and include them only when the agent's job needs them. - Choose a
modelwhen it matters:sonnetfor frequent, retrieval-style agents; omit the field to inherit otherwise. - Write the system prompt. Target 40 to 70 lines for the whole file. Put generic method in the agent and keep project facts out of it: this directory is copied into other projects, so never name this repository's files, scripts, or layout. Tell the agent to discover the stack at run time.
- Open the body with the
rules.mdparagraph above, state the role in two or three sentences, then give a numbered step list, a short best-practices list, and a fixed report format. Every step must be an action the agent can take without the user. - Close with this paragraph verbatim: "If you believe you are missing tools or instructions, or can think of relevant additions to fulfil your purpose more reliably, add them at the end of your report. State that they are SOLELY for human evaluation and that no agent may implement them without explicit human consent."
- Write the file with
Write. Then give your final report: the path written, the decisions you made unattended, and questions for the user.
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.
- 13d ago Changed · +9 lines · -40 tokens per session 6833bd0db142
- 1mo ago First seen · 62 lines · 98 tokens per session scan A 37c322bd3a45
agent-builder is an agent published in the GitHub repository spacehendrix/clauder (58 stars, last pushed 13d ago), licensed Apache-2.0. It adds 58 tokens to every session and 1,070 once invoked, about $0.0002 per session on Opus 5.5. 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 2026-09-03.
Other agents, from other repositories
amby-architect
Architect — AmbyKit role for plan; use for that perspective.
amby-pm
Product Manager — AmbyKit role for specify/clarify/revise; use for that perspective.
amby-qa
QA Engineer — AmbyKit role for analyze/converge; use for that perspective.
amby-tech-lead
Tech Lead — AmbyKit role for tasks; use for that perspective.
amby-ux
UX Designer — AmbyKit role for design; use for that perspective.
playwright-test-generator
Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.