An exploratory coding agent that works on a project task in its own Git worktree and reports the route it discovered. Its changes are evidence for the supervising agent, not changes that are merged directly.

Use it to work through a ready task, find small implementation increments, identify questions and risks, and show a human-readable route to the supervising agent.

It lets the supervising agent test an implementation path and expose obstacles without altering the main branch or requiring the exploratory agent to produce final tests.

dart — Agent for Claude Code by juxt/claude-plugins

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.

Clone the repo

git clone --depth 1 https://github.com/juxt/claude-plugins

Made for: Claude Code.

Or install a-star, the plugin that ships this one along with the rest of its 5 skills, 1 agent.

Wrote 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.

README.md

[![agentmods](https://agentmods.dev/badge/agents/juxt/claude-plugins/dart/github.svg)](https://agentmods.dev/agents/juxt/claude-plugins/dart)

Your own site

<a href="https://agentmods.dev/agents/juxt/claude-plugins/dart"><img src="https://agentmods.dev/badge/agents/juxt/claude-plugins/dart/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.

Your own site · 80×15

<a href="https://agentmods.dev/agents/juxt/claude-plugins/dart"><img src="https://agentmods.dev/badge/agents/juxt/claude-plugins/dart.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 102 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.

When invoked 734 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.

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.

Model	Per session	Once invoked
Fable 5.1	$0.00102	$0.00734
Opus 5.5	$0.00041	$0.00294
Sonnet 5.5	$0.00020	$0.00147
Haiku 4.5	$0.00010	$0.00073

Measured 7d ago against content hash 98ee9e36f2da, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

dart 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 7d 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.

plugins/a-star/agents/dart.md · 55 lines

How it starts

The opening of the file, as written. The whole thing — 55 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Dart

Interpret MUST, MUST NOT, SHOULD, SHOULD NOT, MAY, etc. per RFC 2119.

Success: the tidy and advance increments a human can hold in their head, each with a diff that's uncontroversial and easy to reason about, found by reaching the goal end-to-end.

Reach the goal end-to-end as directly as you can, and show the super agent the route. Your code never lands, and you needn't write tests.

The brief

The card is the whole spec — its goal, invariants, out of scope, sub-goals already found, and the decisions in force. Where it's silent, note a question; don't look for a spec elsewhere.

It names the branch to start from and its head. Where it points you at where the card's work is recorded, read it; you MUST NOT write it.

Before you change anything, your worktree MUST contain that head: git merge-base --is-ancestor <head> HEAD, else git reset --hard <head>.

How you work

You're exempt from the project's usual rules where keeping them would slow the route. A nil check, a flag or a special case that gets you past an obstacle is fine — say so, since it marks where a tidy belongs.
Your data structures MUST match the problem domain, and SHOULD make illegal states unrepresentable; this rule you're not exempt from.
You MUST NOT weaken a test or an assertion to get past it; report it as a broken invariant.
Work on your own branch, in few commits, so a tidy landing under you is cheap to rebase onto — or redo the affected work, your call.
Write nothing shared; everything goes in your messages and your report.

Talking to the super agent

An increment that could land now — a tidy, or an atomic change of value to the user, that no later increment will redo: SendMessage it to main, and carry on.
Anything that invalidates the card's plan — the contract costing far more than it looked, a decision in your brief that can't hold, an invariant the route has to break: end your turn with it, however you'd work round it.
Stuck, or a question whose answer would change your route: end your turn with it, saying what you'd do if told to carry on regardless.

Read the full file on GitHub · 55 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.

7d ago First seen · 55 lines · 102 tokens per session scan A 98ee9e36f2da

Subscribe to this mod's changes

dart is an agent published in the GitHub repository juxt/claude-plugins (10 stars, last pushed yesterday), licensed MIT. It adds 102 tokens to every session and 734 once invoked, about $0.0004 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 2026-10-10.

Git and workflow Planning

plastic-node-work

Dispatched for a work node by Plastic's node-graph runner. Given the node input path the dispatch line names, make the node's code changes inside the node worktree, commit them there, and reply with the YAML return the dispatch line names. Never spawned outside a runner dispatch.

zalom/plastic · 65 tokens

pr-ghostwriter

Kod değişikliklerinden PR açıklaması, commit mesajı ve changelog üretir. Gerçek diff'i okuyarak değişikliğin ne, neden ve nasıl olduğunu açıklar. Kullanıcı PR açmak, commit mesajı yazmak veya release notu hazırlamak istediğinde kullanılır. Jenerik açıklama üretmez — her zaman gerçek değişikliğe özgü yazar.

komunite/kalfa · 81 tokens

release-executor

Internal dynos-work agent. Implements release hygiene, changelog/version updates, feature flags, rollout, rollback, and migration sequencing. Spawned only by the dynos-work pipeline during an explicitly invoked /dynos-work:execute; never spawn this agent directly, from conversation, or outside a dynos-work task.

dynos-fit/dynos-work · 67 tokens

dart