Explore: Add thin orchestration skills for multi-step workflows #71

Closed
opened 2026-07-08 11:54:59 -04:00 by jasper · 0 comments
Contributor

This repo's skills are standalone — each provides a complete capability that an agent loads independently. There's no way to chain them into repeatable multi-step workflows without manual orchestration.

Pocock's skills form a directed graph where orchestrator skills compose primitives. A thin orchestration layer here could chain existing capabilities into repeatable sequences without changing the standalone nature of the individual skills.

What to explore

  1. What workflows naturally emerge from chaining existing skills — e.g. research → distill → vault, or debug → diagnose → fix
  2. Whether orchestrator skills should live in this repo or in a separate workflows repo
  3. How to keep orchestrator skills thin — primarily references to existing skills, not new logic
  4. Whether the agent-skills title/description convention is flexible enough to include process skills alongside capability skills
  5. What the dependency model looks like — orchestrator declares which sub-skills it expects to be available

Candidate workflows

  • research-and-vault: groktocrawl research → extract atoms → create molecule → log session
  • infrastructure-audit: check all running services → compare against expected state → report drift
  • deploy-and-verify: deploy via docker-compose → health check → curl endpoint → report status
This repo's skills are standalone — each provides a complete capability that an agent loads independently. There's no way to chain them into repeatable multi-step workflows without manual orchestration. Pocock's skills form a directed graph where orchestrator skills compose primitives. A thin orchestration layer here could chain existing capabilities into repeatable sequences without changing the standalone nature of the individual skills. ## What to explore 1. What workflows naturally emerge from chaining existing skills — e.g. research → distill → vault, or debug → diagnose → fix 2. Whether orchestrator skills should live in this repo or in a separate workflows repo 3. How to keep orchestrator skills thin — primarily references to existing skills, not new logic 4. Whether the agent-skills title/description convention is flexible enough to include process skills alongside capability skills 5. What the dependency model looks like — orchestrator declares which sub-skills it expects to be available ## Candidate workflows - research-and-vault: groktocrawl research → extract atoms → create molecule → log session - infrastructure-audit: check all running services → compare against expected state → report drift - deploy-and-verify: deploy via docker-compose → health check → curl endpoint → report status
Sign in to join this conversation.
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
magnus/agent-skills#71
No description provided.