agentleFS
lukashouthoofdlukashouthoofd/cenekathon/.claude/skills/ai-in-the-product/SKILL.md

ai-in-the-product

Skill1 starsChanged 15 days agoMITView on GitHub

Use when the product itself calls an LLM or ML model — structured output, prompts, tokens and secrets, quick evals, cost/latency, fallbacks, and the "ML in practice → scalable" story.

  • Reads credentials

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Structured output: ask for JSON matching a schema (Zod/Pydantic), validate, retry once on parse failure.

    line 10
  2. Very specific:

    Test on 5 real examples: before building UI around it.

    line 12
  3. Distinctive:

    No real personal data in prompts.

    line 17

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill 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
SKILL.md

AI inside the product

  • Keys: from .env (.env.example lists the names). Server-side only; never ship a key to the browser or commit it. Organisers provide AI tokens — use the provider/base URL they give.
  • One thin wrapper (src/lib/llm.*) so provider/model swaps are a one-line change. Set timeouts and a cached fallback response for the demo.
  • Structured output: ask for JSON matching a schema (Zod/Pydantic), validate, retry once on parse failure. Don't regex free text.
  • Prompts in files (prompts/*.md), with 3–5 example inputs/outputs beside them.
  • Test on 5 real examples before building UI around it; keep them as fixtures/ — they double as demo data and as your "evaluation" slide.
  • Ground it: feed the model the actual data; show inputs/sources next to answers. Give it an explicit "I don't know" path.
  • Cost/latency: cheapest model that passes your 5 examples; cache by input hash; stream long outputs.
  • Classic ML: baseline (rules/regression) first; report one honest metric. A pre-trained API or model beats training in 4 hours.
  • Pitch hooks: what the model does, what happens when it's wrong, how it scales (batching, caching, queue, eval set), data privacy.
  • No real personal data in prompts; treat model output as untrusted (escape before rendering).

Done when: the AI path works on the 5 fixtures, fails gracefully, and no secret is in git.

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.