Context
Research into agent-skills shows that contextual auto-activation is effective: designing an API automatically triggers the api-and-interface-design skill, building UI triggers frontend-ui-engineering. This minimizes token overhead by loading only relevant context.
Proposal
Add the ability for askcc-cli to auto-detect task type from issue context and load relevant SOPs/skills:
- Parse issue title, body, and labels to classify the task type (API design, frontend, debugging, database, etc.)
- Load only the relevant SOP(s) for that task type into the agent's context
- Fall back to a general-purpose SOP when classification is ambiguous
Rationale
Loading all SOPs into context wastes tokens and dilutes focus. Contextual loading:
- Reduces token usage (progressive disclosure pattern)
- Gives the agent focused, relevant guidance
- Scales as the SOP library grows without increasing per-task overhead
References
- agent-skills: contextual activation based on what the agent is doing
- agent-skills: progressive disclosure — core guidance loads first, supplementary references only when needed
Context
Research into agent-skills shows that contextual auto-activation is effective: designing an API automatically triggers the
api-and-interface-designskill, building UI triggersfrontend-ui-engineering. This minimizes token overhead by loading only relevant context.Proposal
Add the ability for askcc-cli to auto-detect task type from issue context and load relevant SOPs/skills:
Rationale
Loading all SOPs into context wastes tokens and dilutes focus. Contextual loading:
References