Fp minimal

Finish with Proof: a portable execution protocol for coding agents—diagnose before patching, bound delegation, and verify real outcomes.

Install

npx -y skills add MiaoY0uShan/FP --skill fp-minimal

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

Copied from the file, not written here

Minimal FP variant for non-reasoning models. 3 rules, no ceremony.

1.6 KB, as published. Nobody here has run it

The "less is more" variant. 3 rules. ~100 words. For reasoning models.

1. Diagnose before patching. Before changing code, gather evidence to identify the root cause. Do not guess.

2. Verify before claiming done. Never say something is complete without running the relevant tests and observing them pass.

3. Be concise and actionable. First line = result or current action. Last line = next step or final verdict. No preamble, no filler.

That's it. Use your best judgment for everything else.

Reasoning models (gpt-5.6-sol, Claude Opus, etc.) that already have good engineering judgment
Quick tasks where full FP ceremony is overkill
As a baseline to measure whether additional rules actually improve outcomes

In blind evaluations across 1,128 real LLM API calls:

v-minimal (3 rules) scored 3.08 vs full FP (v0) at 3.01 on reasoning models
v-minimal used 5x fewer tokens in multi-turn sessions while maintaining diagnostic accuracy
Both versions correctly refused to guess without evidence — the core behavior is preserved