Home / ericpardee / agent-files · skills/advanced-prompt-improver/SKILL.md · GitHub

advanced-prompt-improver skillA

advanced-prompt-improver is agent-read markdown (skill) from ericpardee/agent-files: Use when improving, reviewing, or writing a prompt for frontier models (Claude 4.6+/Opus 5/Fable 5, GPT-5.x class) - delegation briefs, agent goals, system prompts, CLAUDE.md files, or skills. Triggers on "improve this prompt", "why did this prompt fail", "write the kickoff for", "review my CLAUDE.md/skill", or /advanced-prompt-improver. For quick single-message polish the older prompt-improver also works; this skill is the researched 2026 frontier-model version..

Indexed from public GitHub and served as immutable, content-addressed versions. Install it pinned to an exact SHA-256 with the mdr CLI, and every file is verified against the hash recorded here before it reaches your agent. The deterministic audit below grades the latest version, and the same file always earns the same grade.

What the file says

# Advanced Prompt Improver

Research-grounded (2026 sources at bottom). The one-line thesis, confirmed by Anthropic, OpenAI, and practitioner evidence: frontier models supply the procedure, diligence, and planning by default, so a prompt's job has narrowed to intent, done-state, hard constraints, authorization, and grounding sources, stated exactly once. Specificity did not die; it moved from HOW to WHAT.

## Step 0: Classify before touching anything

- **Conversational or exploratory ask**: leave it alone. Deliberate vagueness is a technique; a vague question gives the model room to surface things you would not ask about.
- **Small task** (diff describable in one sentence): the fix is usually deletion, not addition. No plan ceremony.
- **Delegation or agentic brief** (multi-step, tools, unattended time): apply the six slots below. Full spec in ONE turn; drip-feeding requirements across turns measurably reduces performance.
- **Standing prompt** (system prompt, CLAUDE.md, skill): apply the delete list hardest. Bloated standing instructions cause the model to ignore the rules that matter. Deletion test per line: "would removing this cause a mistake?" If not, cut.
…

Read the whole file at its exact version.

How to install

Latest version
mdr add ericpardee/agent-files/advanced-prompt-improver@git:20260727.291aeae
Exact content
mdr add ericpardee/agent-files/advanced-prompt-improver@sha256:c7d479cfb45c2854

Pin to a label to follow the author's releases, or to a sha256 to freeze the exact bytes forever. Either way the resolved hash is written to mdr.lock, and mdr install reproduces it on any machine.

Badge

mdr badge

[![mdr](https://markdownregistry.com/badge/art_wcdbkdsjhcavctaf.svg)](https://markdownregistry.com/a/art_wcdbkdsjhcavctaf)

1 badge views in 30 days

Versions

versioncommittedcommitsizeaudit
git:20260727.291aeae latest2026-07-27 291aeae 7,274 BA view

Audit of the latest version

A  17 of 17 checks passed. Deterministic, no model, same answer every run.
  • pass: Frontmatter block present
  • pass: Frontmatter declares a name
  • pass: Frontmatter declares a description
  • pass: Size between 200 bytes and 200 KB (7274 bytes)
  • pass: No zero-width or bidi control characters
  • pass: No instruction hidden inside an HTML comment
  • pass: No link to an exfiltration or paste host
  • pass: No credential-shaped string
  • pass: No instruction to send local credentials anywhere
  • pass: No text hidden with inline styles
  • pass: No prompt-injection phrasing
  • pass: No curl or wget piped into a shell
  • pass: No recursive delete of root, home or parent
  • pass: No instruction to read or print local credentials
  • pass: No base64 blob over 200 characters
  • pass: No link to a raw IP address
  • pass: No script tag

Source

GitHub

ericpardee/agent-files · 3 stars · license MIT · pushed 2026-09-20 · branch main

API

GET https://markdownregistry.com/api/v1/artifacts/art_wcdbkdsjhcavctaf
GET https://markdownregistry.com/api/v1/resolve?ref=ericpardee/agent-files/advanced-prompt-improver
GET https://markdownregistry.com/api/v1/blob/c7d479cfb45c2854c23fb2ccc1b5be17f4c12fee8f463742b7b88be96524b969

Your agent does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.

More from ericpardee/agent-files

AGENTS.md agents
ericpardee/agent-files · AGENTS.md
git:20260918.80266df · audit B · 3 stars
CLAUDE.md claude
ericpardee/agent-files · CLAUDE.md
git:20260920.6de3dc1 · audit A · 3 stars
CLAUDE.md@external-memory/templates claude
ericpardee/agent-files · external-memory/templates/CLAUDE.md
git:20260920.afc4daa · audit A · 3 stars
codex-gate skill
ericpardee/agent-files · skills/codex-gate/SKILL.md · Use when User wants a cross-model review gate before shipping code or on a pull request - "before I push/merge/deploy"…
git:20260920.8522c77 · audit A · 3 stars
codex-web-render skill
ericpardee/agent-files · skills/codex-web-render/SKILL.md · Fetch JavaScript-heavy, client-rendered, or bot-protected web pages that the built-in WebFetch tool cannot read…
git:20260918.26378fb · audit A · 3 stars
deliverable-check skill
ericpardee/agent-files · skills/deliverable-check/SKILL.md · Use before any client-facing or stakeholder-facing deliverable ships (report, email draft, Jira/Slack post, doc) -…
git:20260727.291aeae · audit A · 3 stars
freeing-disk-space skill
ericpardee/agent-files · skills/freeing-disk-space/SKILL.md · Use when a Mac is running low on disk space or storage is full - "startup disk almost full", "only X GB left", "free up…
git:20260729.6b547dd · audit A · 3 stars
codex skill
ericpardee/agent-files · skills/skill-codex/SKILL.md · Use when the user asks to run Codex CLI (codex exec, codex resume) or references OpenAI Codex for code analysis…
git:20260920.6de3dc1 · audit A · 3 stars

Every file in ericpardee/agent-files

Browse by kind, by grade A, or by owner.