Entdecke geprüfte X-Beiträge über Skills, die Menschen dahinter und die erwähnten Projekte.
Gezählt werden nur erfasste und geprüfte Beiträge über Skills. Ein leerer Tag bedeutet, dass hier nichts erfasst wurde. Datumsangaben in UTC.
9 erfasste Beiträge
FIRST FLUKE introduces OMA (oh-my-agent): keep Skills once in the repo and link them into Claude Code, Cursor, and Codex so one PR updates all three harnesses.
Abdullah Formuli shares Skill Audit, which profiles agent Skills, flags overlaps and compares competing Skills side by side.
The boss-skill creator explains why copying Skill files does not ensure hooks are connected. The project tracks installed files with a lockfile and checks expected files across target agents with an installation test matrix.
trigger-doctor diagnoses why an existing Skill description misses relevant requests or catches unrelated ones. The creator presents a workflow combining structural checks, simulated trigger judgments, a proposed description and a saved regression suite.
oga_aiichiro ships hotfix v1.0.9 for yomiyasu—the Skill that polishes AI-generated Japanese into more readable, natural prose—so gh skill install works again after a layout fix that keeps a single Skill definition.
Antoine van der Lee shares an audit of his own AGENTS.md and agent sessions: move infrequent instructions out of the default context, remove overlapping SwiftUI skills, and filter noisy build output. He discusses his SwiftUI Skill alongside Caveman and Ponytail, keeping selected ideas rather than claiming that shorter replies always lower total cost.
The creator moved Agent Mission Control into its own Skill folder to reduce what installation copies. He reports a change from 161 files / 3.2 MB to 16 files / 54 KB with the same runtime bytes.
anti-slop v3.2.16 adds `npx antislop-ai --update` to refresh installed Skills. The author notes that plugin installations use separate update steps.
Tw93 discusses keeping Waza’s Skill instructions smaller as models improve. The post names two concrete changes: think compares added complexity with the simplest approach, while check judges whether to keep, adjust or replace an approach before listing patches.