state-of-the-art-first
Skill1 starsChanged 2 months agoMITView on GitHub
Research current techniques when selecting an unfamiliar or fast-changing approach for craft, rendering, audio, or ML.
In this repository16 other files it gives its agents.
- adversarial-refinement-loopSkill
- atomic-commitsSkill
- audio-verificationSkill
- build-without-fearSkill
- comfyui-assetsSkill
- diagnosing-bugsSkill
- finish-the-listSkill
- frontend-designSkill
- headless-terminal-workflowSkill
- parallelize-independent-workSkill
- prefer-second-monitorSkill
- require-external-communication-approvalSkill
- try-it-firstSkill
- visual-verificationSkill
- web-design-guidelinesSkill
- writing-for-agentsSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are 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.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
Start with current primary documentation, maintained implementations, and measurements for the relevant workload.
- Distinctive: line 17
Use enough relevant sources to resolve the decision, without a fixed quota.
≈ “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.
Files
Research the technique that matters
Identify the decision current evidence could change: quality, compatibility, latency, cost, or a known failure mode. Respect the user's chosen tools and constraints. Routine edits using an established technique need no new survey.
Start with current primary documentation, maintained implementations, and measurements for the relevant workload. Add independent comparisons or research when sources disagree or the decision has substantial consequences. Read the material supporting each claim; search snippets are leads, not evidence.
Use enough relevant sources to resolve the decision, without a fixed quota. Expand an explicit broad research brief accordingly. Stop when further sources do not change viable choices, trade-offs, or remaining uncertainty. Keep private project information out of public queries.
Compare candidates under the project's actual hardware, data, and output requirements. A newer model or additional processing stage is a candidate, not an automatic improvement. Distinguish measured results from vendor claims.
For a reversible craft choice, build a small comparison early while filling relevant evidence gaps. Settle expensive or hard-to-reverse dependencies before committing to them. Research should inform the build, not indefinitely delay it.
Record the selected technique, supporting links, trade-offs, and limitations where the project maintains decisions. Verify the claimed benefit on a representative result against the existing or simpler baseline. Reuse that evidence until a new requirement or version invalidates it.
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
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.

