harness:craft-fleet

harness:craft-fleet is a command for Claude Code from Intense-Visions/harness-engineering. It costs 144 tokens per session (966 once invoked), scanned A, original, MIT.

A repository-wide code-quality review that combines multiple focused review skills, ranks their findings, and sends approved work to isolated agents. A worktree is a separate working copy of a Git repository.

In plain words
What is it for?
Use it to review a codebase, group and rank craft issues, approve a batch of findings, split work into fixes or tracked items, and dispatch isolated agents to implement confirmed improvements.
Why use it?
It helps focus quality work on the most important issues while reducing duplicate findings and keeping parallel changes separated.

Command for Claude Code

Part of the harness-claude plugin — 85 skills, 89 commands, 17 agents, 4 hooks, 4 MCP servers shipped together

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 commands/intense-visions/harness-engineering/craft-fleet
Clone the repo
git clone --depth 1 https://github.com/Intense-Visions/harness-engineering

Made for: Claude Code.

Or install harness-claude, the plugin that ships this one along with the rest of its 85 skills, 89 commands, 17 agents, 4 hooks, 4 MCP servers.

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.

agentmods badge for harness:craft-fleet

README.md
[![agentmods](https://agentmods.dev/badge/commands/intense-visions/harness-engineering/craft-fleet.svg)](https://agentmods.dev/commands/intense-visions/harness-engineering/craft-fleet)
Your own site
<a href="https://agentmods.dev/commands/intense-visions/harness-engineering/craft-fleet"><img src="https://agentmods.dev/badge/commands/intense-visions/harness-engineering/craft-fleet.svg" alt="Measured on agentmods" height="20"></a>
Per session 144 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 966 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 $0.00144 $0.00966
Opus 5 $0.00072 $0.00483
Sonnet 5 $0.00029 $0.00193
Haiku 4.5 $0.00014 $0.00097

Measured 5d ago against content hash c86ae95f96b8, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

harness:craft-fleet 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 5d 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-plugin/commands/craft-fleet.md · 41 lines

What it actually says

Phases:

  • select: Compose the enabled craft skills over the repository, fold their findings into (scope, craft domain) targets, apply the noise floor, cross-check each target against open elevation PRs and already-filed quality items, route every surviving finding elevate or file or route by the mechanical boundary and then re-form the targets by verdict so a split scope yields at most one elevate target and at most one file item each carrying only its own findings, enforce the caps in emitted items rather than findings, and order by tier times impact with roadmap-pilot impact scoring
  • confirm: Present the enabled craft domains, the ranked targets with their elevate and file split, a taste-calibration sample of verbatim findings, the noise floor, the per-batch caps, the proposed concurrency, and the pinned base SHA for a single up-front human approval that may disable domains or re-tune the floor
  • dispatch: Fan out worktree-isolated subagents, one per confirmed elevate target, each running the real harness-refactoring pipeline over that target's cited findings and pushing a branch carrying its step-granular commit trail; a target that needs a structural change or whose baseline is not clean at the pinned base — a red suite or a failing harness validate or check-deps — downgrades itself to file and reports, and no subagent runs a re-critique
  • verify: For every elevation independently confirm critique provenance for each changed location, the step-granular refactoring commit trail, and net improvement via a two-run re-critique executed here rather than read from the subagent, plus behavior preservation and all-OS CI green; for every filing, which has no branch, confirm the narrower verified-filing standard of critique provenance plus the cross-check that the target is not already addressed — never by subagent self-report
  • file-and-report: Open one elevation PR per verified target and craft domain without merging, file every verified and every downgraded item as a roadmap item carrying its cite and rubric and location plus an assumptions-made note the orchestrator writes here from the SELECT routing basis, convert every elevation slot to a filed item under --file-only, park routed correctness and security findings for hand-back, and emit a one-row-per-item batch summary including noise-floor drops, over-cap counts, downgrade reasons, cross-check drops, and quiet targets

<execution_context> @agents/skills/claude-code/craft-fleet/SKILL.md @agents/skills/claude-code/craft-fleet/skill.yaml </execution_context>

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. 5d ago First seen · 41 lines · 144 tokens per session scan A c86ae95f96b8

Subscribe to this mod's changes

harness:craft-fleet is a command published in the GitHub repository Intense-Visions/harness-engineering (20 stars, last pushed today), licensed MIT. It adds 144 tokens to every session and 966 once invoked, about $0.0007 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.