anthropics / anthropics/skills

Are external skill contributions welcome? Proposing critical-qa (self-adversarial review before "done")

Aperta
#1,328 2 commenti 0 reazioni 0 assegnatari Vedi su GitHub
Lingua principale
Python
Stelle
176k
Fork
20.9k
Merge medio
7h 21m
PR unite (30g)
5

Descrizione

The skills here (skill-creator, mcp-builder, webapp-testing) read as general tools rather than one-off demos, so I wanted to ask directly: do you accept skill contributions from outside Anthropic, and if so is a PR with a SKILL.md folder following ./template the right path, or do you prefer an issue first?

The skill I would contribute is critical-qa. Before declaring a code change complete, it runs a short self-adversarial pass: it asks hostile-reviewer questions across several fixed categories (mechanism, edge cases, failure modes, confidence, honesty), answers each one honestly, and surfaces the weak spots it found instead of asserting the work is done. It is the quality counterpart to a cost-routing check. In daily use it catches the "looks finished, is not" failure more reliably than a generic review prompt, because the fixed categories force coverage the model would otherwise skip.

It is self-contained: a SKILL.md with the category list and the answer discipline, no external dependencies. If the path is a PR I will follow the template and license it Apache-2.0 to match the repo.

If external skills are not something you are taking right now, no problem, and I would still be glad to know so I can point people to it elsewhere.

Guida per i contributori

Nessuna guida per i contributori indicizzata per questo repository

Direzione di ricerca

This is a policy/proposal question, not a ready task. Read ./template and any contribution guidance in the repository, then wait for a maintainer to say whether external skill folders with SKILL.md are accepted. Done would be a maintainer answer inviting a PR, rejecting the proposal, or documenting the contribution path.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Valutazione

Stack tecnologico
markdown
Ambito
ai-infra-agents, content
Tipo di issue
Funzionalità
Difficoltà
5/5
Tempo stimato
Più di una settimana
Stato di attività
Tranquilla
Chiarezza
Da chiarire
Idoneità per principianti
20/100

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.