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.
npx skills add JairoTorregrosa/jaiskills --skill metapromptgit clone --depth 1 https://github.com/JairoTorregrosa/jaiskillsWrote 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/skills/jairotorregrosa/jaiskills/metaprompt)<a href="https://agentmods.dev/skills/jairotorregrosa/jaiskills/metaprompt"><img src="https://agentmods.dev/badge/skills/jairotorregrosa/jaiskills/metaprompt/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/skills/jairotorregrosa/jaiskills/metaprompt"><img src="https://agentmods.dev/badge/skills/jairotorregrosa/jaiskills/metaprompt.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 | $0.00048 | $0.02374 |
| Opus 5 | $0.00024 | $0.01187 |
| Sonnet 5 | $0.00010 | $0.00475 |
| Haiku 4.5 | $0.00005 | $0.00237 |
Grade A, and why
metaprompt 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 11d 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 — 114 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a prompt engineer. The user provides a goal; you produce a complete prompt optimized for a specific model and harness.
Required inputs
- Goal — what the prompt should accomplish (the user always provides this).
- Target model — which LLM will run the prompt (e.g.,
claude-fable-5,claude-opus-4-8,gpt-5.6-sol,gpt-5.2). If the user omits this, ask: "Which model will run this prompt?" Default toclaude-opus-4-8only if the user says "whatever" or "any". - Target harness — where the prompt will run:
claude-code(CLAUDE.md / skill),codex-cli(AGENTS.md),pi(AGENTS.md / SYSTEM.md),amp(AGENTS.md),eve(agent/instructions.md),ai-gateway(gateway-backed client),api(Messages API / Responses API),chat(conversational UI), oragent-sdk. If the user omits this, ask: "Where will this prompt run — Claude Code, Codex CLI, pi, Amp, eve, AI Gateway, API, chat, or an agent SDK?"
Model → reference routing
| Target model | Reference file |
|---|---|
| Claude Fable 5, Claude Mythos 5 | references/claude-models.md |
| Claude Opus 4.8, 4.7, 4.6 | references/claude-models.md |
| Claude Opus 4.1 (deprecated — retiring Aug 2026; do not target) | references/claude-models.md (migration notes only) |
| Claude Sonnet 5, Claude Sonnet 4.6 | references/claude-models.md |
| Claude Haiku 4.5 | references/claude-models.md |
| GPT-5.6 Sol, GPT-5.6 Terra, GPT-5.6 Luna | references/openai-models.md |
| GPT-5.x (GPT-5, GPT-5.1, GPT-5.2, GPT-5.3) | references/openai-models.md |
| gpt-5-codex | references/openai-models.md |
| Unknown or newer model | Run a freshness web search (see workflow step 2) |
Harness → reference routing
| Target harness | Reference file | Notes |
|---|---|---|
claude-code (CLAUDE.md, skills, subagent prompts) |
references/claude-code-harness.md |
Load alongside the model guide. The harness guide governs what NOT to re-specify (built-in tools, git context, permission system, etc.). |
codex-cli / codex-cloud (AGENTS.md) |
references/codex-harness.md |
Load alongside the model guide. The harness guide governs what NOT to re-specify (sandbox, approval policies, built-in tools, etc.). |
pi |
references/pi-harness.md |
Load alongside the model guide. Pi's ~1k-token system prompt means the generated AGENTS.md (or .pi/SYSTEM.md full replacement) carries more weight than in Claude Code — the guide covers what pi provides (4 core tools, no MCP, no permission rails) and subscription auth (/login OAuth for Claude Pro/Max and ChatGPT). |
amp |
references/amp-harness.md |
Load alongside the model guide — but write the prompt model-agnostically: Amp routes among GLM/GPT/Claude per mode with no model picker, so recommend a mode (low/medium/high/ultra) instead of assuming a model. Default execution is full-auto (no approval gate). |
eve (Vercel agent framework) |
references/eve-harness.md |
The generated prompt is agent/instructions.md (+ optional skill files). Eve is a framework for building agents, not a coding agent; it cannot run on consumer subscriptions — flag API-key/AI Gateway billing. |
ai-gateway |
references/ai-gateway-harness.md |
Not an agent — a model backend. Load together with the actual client harness guide (Claude Code, Codex, eve…). Covers model slugs, routing/fallback options, BYOK, and the subscription-pass-through reality. |
agent-sdk |
references/claude-code-harness.md |
Route to the Claude Code harness guide — section 9 covers headless mode and the Agent SDK specifically. |
api |
Model guide only | No harness reference needed. Output a system prompt + user prompt pair with parameter recommendations. |
chat |
Model guide only | No harness reference needed. Output a self-contained prompt the user can paste into a chat UI. |
What ships with it
8 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 11d ago First seen · 114 lines · 48 tokens per session scan A 8a7ea3db7fbc
metaprompt is a skill published in the GitHub repository JairoTorregrosa/jaiskills (5 stars, last pushed 10d ago), licensed MIT. It adds 48 tokens to every session and 2,374 once invoked, about $0.0002 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.
Other skills, from other repositories
prompt-engineering
Provides workflows to write, debug, and optimize prompts for LLMs, including few-shot example selection, chain-of-thought structuring, system prompt design, and template composition. Use when the user asks to write or improve a prompt, wants help with few-shot examples, chain-of-thought, system prompts, prompt…
langchain4j-tool-function-calling-patterns
Provides and generates LangChain4j tool and function calling patterns: annotates methods as tools with @Tool, configures tool executors, registers tools with AiServices, validates tool parameters, and handles tool execution errors. Use when building AI agents that call tools, define function specifications, manage…
loom-prompt-engineering
Designs and optimizes prompts for large language models including system prompts, agent signals, and few-shot examples.
improve-prompt
Critique and rewrite a prompt using prompt engineering best practices: clarity, examples, XML structure, role, explicit output format, and positive-over-negative instructions. Asks 1-3 targeted questions to fill missing context, then returns a short critique plus a drop-in rewritten prompt. Use when the user wants to…
write-a-prompt
Creates a copy-ready prompt from a rough request, notes, source material, or the current conversation using OpenAI's prompting guidance. Use when the user invokes $write-a-prompt or /write-a-prompt, types a common misspelling such as /write-a-promopt or /wite-a-prompt, asks to "write a prompt for me," asks to turn the…
prompt-evaluation-runner
Use when evaluating prompts, LLM outputs, red-team suites, or model behavior with local eval configs and safe provider/cost controls.