ship CLAUDE.md

ship CLAUDE.md is an instructions file for coding agents from kronael/ship. It costs 1,595 tokens per session, scanned A, original, MIT.

A safety checklist for language-model features, meaning software that uses an AI model to produce or act on text. It examines prompt injection, made-up answers, unsafe output use, excessive agent permissions, and abuse cases.

In plain words
What is it for?
Use it before shipping user-facing AI features, accepting user input or uploads into prompts, or releasing agents that write files, call services, or send messages.
Why use it?
It helps find ways users or uploaded content could manipulate the model or cause harmful behavior before the feature is released.

Instructions file

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 instructions/kronael/ship/claude-md
Clone the repo
git clone --depth 1 https://github.com/kronael/ship

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 ship CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/kronael/ship/claude-md.svg)](https://agentmods.dev/instructions/kronael/ship/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/kronael/ship/claude-md"><img src="https://agentmods.dev/badge/instructions/kronael/ship/claude-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 1,595 This file is loaded in full into every session.
When invoked 1,595 The same file — it is already loaded in full.
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 $0.01595 $0.01595
Opus 5 $0.00797 $0.00797
Sonnet 5 $0.00319 $0.00319
Haiku 4.5 $0.00160 $0.00160

Measured 3d ago against content hash 23c0f487d8e3, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

ship CLAUDE.md 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 3d 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.md · 121 lines

How it starts

The opening of the file, as written. The whole thing — 121 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

autonomous agent orchestration. planner-worker-judge pattern.

build and run

make build              # uv sync
make install            # uv tool install + skill
make test               # pytest
make lint               # pre-commit run -a
make right              # pyright only
make clean              # rm cache + state
ship <file>             # ship from file
ship <dir>              # ship from dir
ship <arg> <arg> ...    # args as context
ship -f                 # wipe state, start fresh
ship -k                 # validate spec only (exit 0/1)
ship -n 8 -t 600 -m 10 # override workers/timeout/turns
ship -s                 # skip validation, trust spec as-is
ship -p "use stdlib"    # inject override into all LLM calls
ship -x                 # enable refiner (codex CLI critique)
ship -v                 # verbose (-v: details, -vv: debug)
ship -q                 # quiet (errors only)
ship -h                 # help

defaults

  • workers: 4
  • max_turns: 50
  • task_timeout: 2400s (40min, agents told the real value)
  • spec discovery: file or dir required on cmdline (no auto-discovery)
  • verbosity: 1 (0=quiet, 1=default, 2=verbose, 3=debug)

how it runs

  1. validator checks spec -> rejects to .ship/REJECTION.md or writes PROJECT.md; result cached in .ship/validated (SHA256)
  2. planner runs once -> breaks design into tasks (with mode + worker assignment)
  3. workers pull from queue, execute via claude CLI (parallel or sequential)
  4. judge polls every 5s, updates TUI panel, retries failed tasks
  5. when all complete: refiner (codex CLI) critiques batch if -x enabled, creates follow-up tasks
  6. if refiner finds nothing: replanner (claude CLI) does full assessment vs goal
  7. if replanner exhausted: adversarial verifier generates challenges as tasks (3 rounds max)
  8. loop continues until no new tasks generated

critical patterns

if task.status is TaskStatus.PENDING:  # use 'is' not ==

Read the full file on GitHub · 121 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. 3d ago First seen · 121 lines · 1,595 tokens per session scan A 23c0f487d8e3

Subscribe to this mod's changes

ship CLAUDE.md is an instructions file published in the GitHub repository kronael/ship (1 stars, last pushed 6mo ago), licensed MIT. It adds 1,595 tokens to every session, about $0.0080 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 2026-08-31.