prompt-design

prompt-design is a skill for Claude Code, Codex from fmind/dot. It costs 43 tokens per session (1,347 once invoked), scanned A, original, MIT.

A method for designing instructions for language models and agent applications. It covers how instructions, context, tools, retrieved information, and structured outputs fit together and which rules take priority.

In plain words
What is it for?
Use it to write production prompts, define tool contracts, manage retrieved context, and record the exact setup behind a prompt version.
Why use it?
It helps prevent unclear instructions, conflicting context, unsafe tool behavior, and unrepeatable prompt changes. It also makes proposed prompts easier to test and compare.

Skill for Claude CodeCodex

Which agent this was written for is unclear — built for hermes-agent. Also seen: built for hermes-agent.

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 skills/fmind/dot/prompt-design
Any agent
npx skills add fmind/dot --skill prompt-design
Clone the repo
git clone --depth 1 https://github.com/fmind/dot

Made for: Claude Code, Codex.

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 prompt-design

README.md
[![agentmods](https://agentmods.dev/badge/skills/fmind/dot/prompt-design.svg)](https://agentmods.dev/skills/fmind/dot/prompt-design)
Your own site
<a href="https://agentmods.dev/skills/fmind/dot/prompt-design"><img src="https://agentmods.dev/badge/skills/fmind/dot/prompt-design.svg" alt="Measured on agentmods" height="20"></a>
Per session 43 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,347 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 5.1 $0.00043 $0.01347
Opus 5 $0.00022 $0.00674
Sonnet 5 $0.00009 $0.00269
Haiku 4.5 $0.00004 $0.00135

Measured 2d ago against content hash e16cd63d2074, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-05, from the pricing page.

Security

Grade A, and why

prompt-design 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.

skills/prompt-design/SKILL.md · 45 lines

How it starts

The opening of the file, as written. The whole thing — 45 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Prompt Design

Turn a behavioral contract into the smallest production prompt stack that expresses it, then hand a frozen candidate to agent-evaluation, which owns proof of behavior. agent-skills owns reusable SKILL.md packages and agent-project repository-level instructions.

Workflow

  1. Name the success contract: user-visible behavior, failure semantics, forbidden actions, measurable acceptance criteria, and the cheapest development cases that separate success from failure; product ambiguity goes to product-loop first.
  2. Inspect the runtime input: trace every layer in precedence order (provider rules, system and developer text, tenant customization, memory, tool schemas, retrieval, history, user input), note truncation and caching, and design against the runtime prompt, not a file.
  3. Pin the baseline: code revision, prompt hash, assembly implementation, model version, tool and output schemas, retrieval snapshot, context limit, sampling, retries, and known results; separate prompt changes from system changes.
  4. Partition context by lifetime: stable policy and tool contracts early and cacheable, tenant or session context in a bounded layer, volatile request state last; define precedence and deterministic truncation before the window fills.
  5. Write one behavioral contract: each instruction in one authoritative place, stating goal, constraints, decision authority, success criteria, failure behavior, and output contract; cut persona flourishes, repeated rules, and speculative edge cases.
  6. Encode authority and autonomy: what the agent may read, write, call, spend, send, or publish, which actions need confirmation, and when it must stop; conflicts resolve by explicit priority, never by recency or persuasive wording inside data.
  7. Design tool contracts: unique action-oriented names, a concise purpose, and the field list in tool contracts; expose only the tools relevant to the task.
  8. Make outputs machine-checkable: a typed schema or discriminated result variants with required fields, enums, nullability, evidence fields, and refusal or partial-success shapes; validate in code and fail closed.
  9. Choose examples at decision boundaries: the fewest examples that resolve an ambiguous rule, output shape, tool choice, or refusal, including hard negatives; never copy sealed evaluation cases into the prompt.
  10. Harden dynamic insertion: typed template parameters, explicit delimiters, length bounds, format-appropriate escaping, provenance labels, and deterministic placement; reject missing variables instead of emitting placeholders.
  11. Run static checks: render the candidate with representative values and inspect it in final order for contradictions, unknown tools or fields, schema-invalid examples, unresolved variables, authority inversion, and rules the runtime cannot enforce.
  12. Hand off to evaluation: diff and hash baseline and candidate, state one change hypothesis with its guardrails, freeze the candidate, and deliver the prompt candidate to agent-evaluation for paired trials.

Read the full file on GitHub · 45 lines

Files

What ships with it

2 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.

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 · 45 lines · 43 tokens per session scan A e16cd63d2074

Subscribe to this mod's changes

prompt-design is a skill published in the GitHub repository fmind/dot (4 stars, last pushed yesterday), licensed MIT. It adds 43 tokens to every session and 1,347 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 2026-09-03.

Related

Other skills, from other repositories