meta-prompt

A prompt-writing command that turns a rough request or conversation into a complete brief for another AI coding session.

In plain words
What is it for?
Use it to prepare copy-and-paste prompts, handoffs, or instructions for another agent without asking that agent to perform the current task.
Why use it?
It carries over relevant decisions and constraints so the receiving agent can understand the task without seeing the earlier conversation.

Command

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/bengous/claude-code-plugins/meta-prompt
Clone the repo
git clone --depth 1 https://github.com/bengous/claude-code-plugins
Per session 87 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,383 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 $0.00087 $0.01383
Opus 5 $0.00044 $0.00691
Sonnet 5 $0.00017 $0.00277
Haiku 4.5 $0.00009 $0.00138

Measured 2d ago against content hash 398bceb04883, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

meta-prompt 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-meta-tools/commands/meta-prompt.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.

Meta Prompt Enhancer

Turn the user's rough request into a prompt another agent can execute without further context.

Input

$ARGUMENTS

No arguments means the request is the task discussed in this conversation — write the prompt that hands it off.

Order of work

Mine the conversation when there is one, ground the request, clarify only if that left a blocking gap, settle who will run it, then emit the prompt in the shape described under Output contract at the end of this file. Do not execute the task you are writing the prompt for.

Mine the conversation

When the request follows a discussion — planning, debugging, a design back-and-forth — the conversation holds facts grounding cannot recover: decisions made and the why behind them, constraints and anti-patterns the user expressed, prior art they pointed at. Carry those into the prompt; the why matters because the executor uses it for judgment calls the prompt cannot anticipate. Filter ruthlessly — most of the conversation is irrelevant to the executor.

The executor reads cold. No shorthand from this conversation ("as discussed", "the earlier approach") — every reference must resolve for a reader with zero context.

Ground the request first

A handoff prompt that cites paths, functions, or conventions that do not exist is worse than a vague one — the receiving agent trusts it and acts on it. So verify before asserting:

  • Confirm any file, symbol, or pattern the request names actually exists, with Grep/Glob/Read.
  • Pick up the project's instruction files (CLAUDE.md, AGENTS.md, .claude/rules/) and carry over the conventions bearing on this task. Check what the harness already put in context before re-reading them.
  • Cite real paths and real line numbers. Where you could not verify something, label it unverified in the prompt rather than asserting it.

When grounding contradicts the request, grounding wins. If the thing already exists, is already partly built, is named differently, or does not exist at all, retarget the prompt at the true state and say so in your rationale line. Name anything you deliberately left out of scope. Retargeting to fit reality is not scope creep; writing a prompt for work that is already done is the actual failure.

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 · 87 tokens per session scan A 398bceb04883

Subscribe to this mod's changes

meta-prompt is a command published in the GitHub repository bengous/claude-code-plugins (4 stars, last pushed 3d ago), licensed MIT. It adds 87 tokens to every session and 1,383 once invoked, about $0.0004 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.