Small skill
Refine an Agent Skill via the skill-forge judge → hitl loop, in a sandboxed Docker container. npx-installable.
npx -y skills add barryroodt/refine-skill --skill small-skillAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 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.
What its author says it does
Copied from the file, not written here
Minimal skill used to exercise the refine loop end-to-end with a real model.
SKILL.md
0.4 KB, 67 tokens by cl100k_base, as published. Nobody here has run it
Goal
Demonstrate that the refine loop runs against a real LLM.
Process
- Read the SKILL.md and judge it.
- If improvable, apply one or two changes.
- Stop when no meaningful improvements remain.
Anti-patterns
- Don't add features unrelated to the goal.
Gives 0 of the 12 instructions most e2e browser skills give in 67 tokens
Counted across 407 of the 410 authors here whose files we hold, read 2026-08-06
- use page object model patternin 35 of 407, across 25 files
- Snapshot to get element refsin 24 of 407, across 14 files
- keep tests independentin 23 of 407, across 18 files
- Interact using refs from the latest snapshotin 23 of 407, across 11 files
- clean up test data after each testin 21 of 407, across 15 files
- test user behavior not implementationin 20 of 407, across 14 files
- quarantine flaky tests explicitlyin 19 of 407, across 10 files
- wait for specific network conditionsin 18 of 407, across 8 files
- re-snapshot after navigation or dom changesin 17 of 407, across 10 files
- Detect running dev servers before writing test codein 17 of 407, across 7 files
- use web-first assertionsin 17 of 407, across 14 files
- capture screenshots or videos on test failurein 17 of 407, across 14 files
Said here and by no other author read
- read the skill
- judge the skill
- apply one or two improvements if possible
- stop when no meaningful improvements remain
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.