Nothing matches — try another term.
The skills people actually install — drawn from skills.sh and ranked by popularity — each one compiled into a multi-agent workflow: independent steps run in parallel, every finding is adversarially verified, and the hand-offs are typed. Same procedure, a fraction of the wall-clock.
Each skill is read as a procedure, decomposed into its true dependency graph, and re-emitted as orchestration — then checked before it ships.
Read the SKILL.md and every bundled reference and script.
Classify each step: sequential, parallel, per-item, or a finding to verify.
Emit a runnable workflow — fan-out, pipeline, adversarial checks, typed hand-offs.
Parse-gate the result and review it before it enters the index.
The compiler is open source — skill-to-workflow. Compile your own with npx github:democra-ai/skill-to-workflow