← All popular talks

Popular talk #30

Building Great Agent Skills: The Missing Manual

Synced transcript

Follow the talk

Community discussion

Add context, evidence, or a useful disagreement

Build on the talk with an example, caveat, connection, or elaboration. Draft here, add the moment you’re discussing, then choose the direct-post pilot or the YouTube handoff.

Make it useful: add one concrete point, then support it with context, evidence, an example, or a caveat. Your words stay exactly as written.

Automated overview

What this talk covers

Matt Pocock presents a 'skill checklist' framework to help developers escape 'skill hell' by systematically evaluating and improving agent skills, covering trigger, structure, steering, and pruning. He explains the trade-offs between user-invoked and model-invoked skills: model-invoked increase context load and unpredictability, user-invoked increase pilot cognitive load. Structure divides skills into steps and reference, with branching material hidden behind context pointers to minimize skill.md. Steering uses 'leading words' like 'vertical slice' that appear in reasoning traces; to boost leg work, he splits skills so agents only see one step at a time. Pruning removes duplication, sediment, and no-ops that don't affect behavior. The framework is in the 'writing-great-skills' skill in his repository.

This overview is derived from the transcript and has not been independently fact-checked by AI Engineer.

Generated highlights

Key moments

  1. 0:40insightSkill Hell

    Skill hell is the inability to distinguish good AI agent skills from bad ones, preventing effective use and organization adoption.

  2. 5:17insightTrigger

    Adding a model-invoked skill to an agent increases context load by permanently inserting its description into the context window.

  3. 7:29insightStructure

    Skills are built from two units: steps (the procedure) and reference (supporting information), simplifying skill design.

  4. 12:23insightSteering

    Leading words pack dense meaning into short phrases, guiding agent reasoning traces when repeated in the skill and thinking tokens.

  5. 18:31insightPruning

    No-ops are skill instructions that appear useful but do not change agent behavior, such as asking for a long commit message when it's already the default.

  6. 19:55insightClosing

    The 'writing great skills' skill encodes the entire checklist framework, available on Matt Pocock's GitHub for immediate use.

These moment labels are generated from the unreviewed transcript and may contain errors.