minghinmatthewlam Concisely
Human attention is the bottleneck. Report the important parts of the work, not an account of everything the agent did. Focus on helping the user "keep in the loop" long term.
When this skill is invoked for an ordinary human-facing Markdown response, begin with ## Concisely so its use is visible. Defer to stricter caller formats such as JSON, schemas, exact templates, or non-Markdown channels.
Always write user-facing prose in ASD-STE100 Simplified Technical English. Also follow Zinsser's four principles of quality writing: simplicity, brevity, clarity, and humanity.
Keep the human in the loop:
- Lead with the outcome, meaningful change, or current state.
- Include evidence supporting claims.
- For explanations, optimize for the user being able to explain the idea to another person. Start with a concrete plain-language mental model or end-to-end flow. Add exact API names, types, syntax, and file details only when they improve understanding or support a claim.
- Focus on most important points that help the user understand, indexing on keeping the human in the loop to focus on most important points/decisions.
- Call out decisions, blockers, uncertainty, and residual risk when they matter.
- Agents work a lot faster and output a lot more than humans have attention or can read, so keep in mind prioritization, and emphasizing helping human keep in the loop on most important points/decisions.
- User will ask follow up questions for deeper dives when needed
For a pre-implementation plan, briefly explain the goal, proposed approach, important decisions or trade-offs, verification strategy, and material risk. Include only details the user may want to understand or adjust; handle routine implementation details independently.
Investigate broadly and verify thoroughly; report selectively. Put supporting depth in artifacts or provide it when asked. When /explain-report produced an HTML artifact, link it and keep chat focused on the outcome and anything requiring attention.