agentleFS
Sign inStart free
mehbylabsmehbylabs/creator-skills/AGENTS.md

creator-skills

AGENTS.md7 starsChanged 16 days agoMITView on GitHub

Provider-agnostic creator skills: one skill per creator job, any backend the user has. Never edit synced copies in skills/*/scripts/ or skills/*/references/ when the file exists in shared/; edit shared/ and run scripts/sync-shared.sh. Skill-specific references (e.g. creator-product-photo/references/modes.md) live only in their skill. 1. Detect, don't assume. No skill hard-codes a provider; all generation goes through creator.py, which honors user config, then profile defaults. 2. Scriptable before manual. API/CLI providers before computer-use ones. 3. Fallback never bypasses safety. A policy refusal stops…

The file

We split this AGENTS.md into the rules it gives an agent

6 rules. About 0 are common, 0 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Never edit synced copies in skills/*/scripts/ or skills/*/references/ when the file exists in shared/.

    line 20
  2. Very specific:

    Detect, don't assume: No skill hard-codes a provider.

    line 26
  3. Very specific:

    No other top-level keys. Keep SKILL.md under 300 lines.

    line 51
  4. Very specific:

    Add to DEFAULT_ORDER (both profiles) and DEFAULT_MODELS.

    line 57
  5. Very specific:

    Document in shared/references/providers.md.

    line 58
  6. Very specific:

    Create skills/creator-<job>/SKILL.md, run sync + validate, add it to README and skills.sh.json.

    line 63

≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
AGENTS.md

AGENTS.md — creator-skills

Provider-agnostic creator skills: one skill per creator job, any backend the user has.

Layout

shared/                   # canonical source — edit here only
  scripts/creator.py      # detect / plan / image / video / speech / music / sfx / adopt
  scripts/media.py        # local ffmpeg ops: probe / still / text / cutout / reframe / captions / assemble
  references/*.md         # providers, config, prompting, audio, media ops, browser playbooks
skills/<name>/            # installable skills (discovered by `npx skills`)
  SKILL.md
  scripts/, references/   # synced copies of shared/ + skill-specific references
scripts/sync-shared.sh    # copy shared/ into every skill (--check for CI)
scripts/validate.py       # spec + convention checks (run before every commit)
skills.sh.json            # grouping on the skills.sh repo page

Never edit synced copies in skills/*/scripts/ or skills/*/references/ when the file exists in shared/; edit shared/ and run scripts/sync-shared.sh. Skill-specific references (e.g. creator-product-photo/references/modes.md) live only in their skill.

Principles

  1. Detect, don't assume. No skill hard-codes a provider; all generation goes through creator.py, which honors user config, then profile defaults.
  2. Scriptable before manual. API/CLI providers before computer-use ones.
  3. Fallback never bypasses safety. A policy refusal stops the chain.
  4. Verify on disk. A provider's success message is not proof; the file is.
  5. The host agent orchestrates. agy/codex are invoked as providers only through creator.py; the host writes prompts and reviews output.
  6. Stdlib only. Scripts must run on a bare Python 3.9+ with ffmpeg.

SKILL.md frontmatter (Agent Skills spec)

---
name: creator-<job>            # == folder name, [a-z0-9-], ≤ 64
description: >-                # ≤ 1024 chars: what it does, "Use when …", "Not for …"
  …
license: MIT
compatibility: …               # ≤ 500 chars
allowed-tools: Bash Read Write # space-separated string
metadata:
  version: "0.3.0"             # must equal VERSION
  argument-hint: "…"
---

No other top-level keys. Keep SKILL.md under 300 lines; move tables/examples to references.

Adding a provider

  1. PROVIDERS entry (per-task capabilities) and detect_one branch in shared/scripts/creator.py.
  2. Adapter functions raising ProviderError(kind, msg); register in ADAPTERS.
  3. Add to DEFAULT_ORDER (both profiles) and DEFAULT_MODELS.
  4. Document in shared/references/providers.md; browser providers get browser-<id>.md.
  5. scripts/sync-shared.sh && python3 scripts/validate.py.

Adding a skill

Create skills/creator-<job>/SKILL.md, run sync + validate, add it to README and skills.sh.json.

Versions

VERSION is the source of truth; every metadata.version must match. Bump all together.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free