Building Great Agent Skills: The Missing Manual
summary
Gives a concrete checklist: pick user-invoked or model-invoked and accept the tradeoff between context tokens versus pilot cognitive load, split skills into steps and reference and move branch-only templates out of the main skill.md so the skill.md minimal rule saves tokens and maintenance, use repeatable leading words (e.g. “vertical slice”) to change reasoning traces, split phases into separate skills (example: grill with docs then 2PRD) and prune sediment, duplicates and no‑ops until the skill is small and deterministic.
AI-written summary · about
chapters (10)
- 0:00Introduction to the talk and the concept of "skill hell"
- 2:12Overview of the skill checklist framework
- 3:16Trigger: Choosing between user-invoked and model-invoked skills
- 7:29Structure: Organizing steps and references
- 9:00Making the skill.md file minimal
- 11:54Steering: Using leading words to guide agent behavior
- 14:56Increasing "leg work" per step
- 16:48Pruning: Removing sediment, crud, and no-ops
- 19:06Final summary of the checklist framework
- 19:55Where to find the "writing great skills" resource
related talks
29mStop Prompting, Start Composing: Agent Skills in Practice
35mSkills.md and agents.md You Must Master in 2026!
01h 21mFull Walkthrough: Writing & Using Skills
31mSpecialize Your Agents with Skills
21mBest practices are trapped in your head and your AI has no idea
01mSpecialize Your Agents with Skills
18m"Software Fundamentals Matter More Than Ever"
01h 19mSkill Issue: How We Used AI to Make Agents Actually Good at Supabase