Videos UNzCG3lw6O0

Building Great Agent Skills: The Missing Manual

index_state ready data_status ok

AI Engineer· published 2026-06-29· 0:20:43· en-US· indexed 2026-08-10 19:46

Open on YouTube

56 shot(s).

keyframes kept every frame deduplicated

What was stored

cues: 245; whisperx 245
chunks: 37; from 245 cues
keyframes: 55; kept of 56 captured
frames with text: 55; 191 lines read
chapters: 10; from the source metadata
keyframe bytes: 4.2 MB; word timings on 245 cues

Provenance

Each pipeline stage, its state and the model that produced it
stage	state	model	started	took
`fetch`	done	—	2026-08-10 01:28	2m 56s
`stt`	done	—	2026-08-10 01:31	24s
`chunk`	done	—	2026-08-10 01:31	0s
`text_embed`	done	—	2026-08-10 19:45	0s
`keyframe`	done	—	2026-08-10 01:31	2m 04s
`ocr`	done	—	2026-08-10 01:33	10s
`frame_embed`	done	—	2026-08-10 19:45	10s

Frames, and what the machine read

0:08 #0 done1 line(s)

shot 0·sharpness 219.9
1. GRASS0.99
0:25 #1 done4 line(s)

shot 1·sharpness 814.8
1. The Missing1.00
2. Manua10.92
3. How To Write Great Skills1.00
4. Matt Pocock0.99
0:43 #2 done1 line(s)

shot 2·sharpness 472.0
1. Framework Hell0.99
1:03 #3 done1 line(s)

shot 3·sharpness 375.5
1. Skill Hell0.97
1:36 #4 done2 line(s)

shot 4·sharpness 628.1
1. just one more1.00
2. skill bro0.99
1:42 #5 done1 line(s)

shot 5·sharpness 534.3
1. mattpocock/skills0.99
2:01 #6 done3 line(s)

shot 6·sharpness 708.3
1. What makes a1.00
2. skill0.95
3. great?1.00
2:11 #7 done1 line(s)

shot 7·sharpness 500.1
1. A Skill Checklist0.99
2:28 #8 done4 line(s)

shot 8·sharpness 994.2
1. 1. Trigger1.00
2. 2. Structure1.00
3. 3. Steering1.00
4. 4. Pruning0.98
3:13 #9 done1 line(s)

shot 9·sharpness 517.6
1. /writing-great-skills1.00
3:36 #10 done3 line(s)

shot 10·sharpness 1025.0
1. mattpocock/skills1.00
2. VS0.86
3. obra/superpowers1.00
3:49 #11 done1 line(s)

shot 11·sharpness 341.9
1. /my-skil10.94
4:11 #12 done4 line(s)

shot 12·sharpness 661.0
1. Description1.00
2. SKILL.md1.00
3. >0.92
4. 10.53
4:41 #13 done1 line(s)

shot 13·sharpness 455.0
1. Context Pointer1.00
4:50 #14 done3 line(s)

shot 14·sharpness 732.9
1. Context Pointer1.00
2. (optional)1.00
3. R0.51
4:54 #15 done34 line(s)

shot 15·sharpness 3220.3
1. SKILL.md.../grill-with-docs0.98
2. SKILL.md.../domain-modeling0.98
3. SKILL.md.../codebase-design X0.99
4. SKILL.md.../grill-me0.97
5. Q0.96
6. skills>engineering>codebase-design>SKILL.md0.98
7. 11.00
8. 21.00
9. name:codebase-design0.99
10. 31.00
11. description: Shared vocabulary for designing deep modules. Use1.00
12. when the user wants to design or improve a module's interface,1.00
13. find deepening opportunities, decide where a seam goes, make1.00
14. code more testable or AI-navigable, or when another skill needs1.00
15. the deep-module vocabulary.0.99
16. 51.00
17. # Codebase Design1.00
18. 81.00
19. Design **deep modules**: a lot of behaviour behind a small1.00
20. interface, placed at a clean seam, testable through that0.99
21. interface. Use this language and these principles wherever code1.00
22. is being designed or restructured. The aim is leverage for1.00
23. callers, locality for maintainers, and testability for everyone.1.00
24. 91.00
25. 101.00
26. ## Glossary0.98
27. 111.00
28. 121.00
29. Use these terms exactly - don't substitute "component," "service,0.99
30. " "API," or "boundary." Consistent language is the whole point.0.99
31. 131.00
32. 141.00
33. **Module** - anything with an interface and an implementation.0.99
34. Deliberately scale-agnostic: a function, class, package, or0.99
5:00 #16 done18 line(s)

shot 16·sharpness 693.2
1. SKILL.md .../grill-with-docs0.97
2. SKILL.md.../domain-modeling0.99
3. SKILL.md .../codebase-design0.97
4. SKILL.md .../grill-me × Q0.90
5. skills> productivity>grill-me>SKILL.md0.96
6. 11.00
7. 21.00
8. name: grill-me0.99
9. 31.00
10. description: A relentless interview to sharpen a plan or design.0.99
11. 41.00
12. disable-model-invocation: truel0.99
13. I0.90
14. 51.00
15. 60.89
16. 70.99
17. Run a /grillingsession.1.00
18. 81.00
5:10 #17 done4 line(s)

shot 17·sharpness 1146.7
1. Tip #10.95
2. Decide if your skill1.00
3. is user-invoked or1.00
4. model-invoked1.00
5:57 #18 done3 line(s)

shot 18·sharpness 703.2
1. User-invoked1.00
2. Cognitive Load1.00
3. L0.64
6:22 #19 done3 line(s)

shot 19·sharpness 1034.8
1. mattpocock/skills1.00
2. VS0.86
3. obra/superpowers1.00
6:54 #20 done3 line(s)

shot 20·sharpness 807.4
1. Context Pointer1.00
2. Unpredictability1.00
3. >0.86
7:21 #21 done4 line(s)

shot 21·sharpness 1143.5
1. Tip #10.96
2. Decide if your skill1.00
3. is user-invoked or0.98
4. model-invoked1.00
7:33 #22 done1 line(s)

shot 22·sharpness 345.4
1. 2. Structure0.99
7:43 #23 done4 line(s)

shot 23·sharpness 1054.3
1. Steps1.00
2. The step-by-step procedure1.00
3. Reference1.00
4. Supporting information1.00

Transcript

245 cues· 4,183 words· 22,443 chars

chunk 0 · 0:00–0:39 · 146 words · 764 chars
0:00 Hello, friends.
0:00 I was dearly hoping to be able to come to the AI Engineer World's Fair, but family matters have intruded and I'm not able to make it.
0:07 However, I will not be leaving you empty-handed.
0:09 I'm going to give you the talk that I would have given in San Francisco.
0:13 This talk is called The Missing Manual, How to Write Great Skills.
0:17 And
0:17 I think that the ability to distinguish good skills from bad skills is only getting more important.
0:22 As developers, we seem to be pretty talented at finding different forms of hell for us to go to.
0:28 In, like, a few years ago, we had tutorial hell, which is where you would go into a bunch of tutorials trying to learn something, not be able to piece it together and sort of just get into this cycle you couldn't get out of.
chunk 1 · 0:40–1:19 · 144 words · 791 chars
0:40 We had framework hell, where every other 10 minutes there was a JavaScript framework being announced, and you had to learn the hot new thing all the time.
0:48 And now I think we have another version of hell, which is skill hell.
0:51 Skill hell is where you have all of these skills available, freely available, that you can download, contribute to, you can figure out on your own, but you don't really know how the pieces all work together.
1:01 You can't tell a good skill from a bad skill.
1:04 And this means that people are trying to piece together these frameworks, trying to
1:07 try everything that's out there all at once.
1:10 And they sort of can't, or rather, they don't get the results that the skills themselves promise.
chunk 2 · 1:15–1:58 · 163 words · 820 chars
1:15 This is true at an individual level, but it's also true at an organization level too.
1:19 Organizations have no way or no understanding on how to build good skills, how to take their operating procedures and turn them into things that an agent can do.
1:28 And if you don't do that, then it's hard to get the bounty that skills can offer.
1:33 Just one more skill, bro.
1:35 That kind of seems like what we're saying.
1:37 And I feel a bit of guilt here too, because we have Matt Pocock skills, which is my skills repo, which is one of the most popular engineering skill sets out there.
chunk 3 · 1:45–2:28 · 172 words · 850 chars
1:45 And so I feel like I want to help the people who use my skills get out of skill hell.
1:50 So how do we do it?
1:50 How do we get out of this?
1:51 Well, what is actually missing?
1:54 Well, in my opinion, the thing that we're missing is we don't know what makes a skill great.
1:58 We can't yet look at a skill and go, okay, this skill is doing these good things and these bad things.
2:04 There's no shared rubric, no framework for looking at a skill and making it better.
2:08 And so that's what I'm going to give you in this talk.
2:10 I'm going to give you a skill checklist, a checklist of things you can look at inside the skill to make sure that it's doing what it says it's doing and ways you can improve it, ways you can write skills.
chunk 4 · 2:20–3:02 · 149 words · 781 chars
2:20 This checklist looks like this.
2:22 We start with the trigger of the skill, how the skill is invoked and the decisions that you need to design there.
2:29 Then the internal structure of the skill, how the skill is actually composed and laid out internally.
2:35 Then number three is how do you actually steer using the skill?
2:38 How do you get the skill to tell the agent what to do?
2:42 And then four, how do you make the skill as small as possible?
2:46 Because once we've got a working skill, we then need to basically maximize it, prune out all of the irrelevant stuff, prune out all of the no-ops.
chunk 5 · 2:53–3:34 · 155 words · 850 chars
2:53 There's one handy advantage of me not being in the room with you, which is you can immediately go and try this out because I've encoded all of this into a new skill in my repo called writing great skills.
3:03 So if you've got an immediate use case for this, just go to this MySkills repo, you know, just close this browser, get out of here and go and use this skill to either improve your skills or write great new ones.
3:14 But let's start going through the checklist then.
3:15 We have number one, the trigger, the way the skill is invoked.
3:19 And in order to talk about this, I'm actually going to do a bit of comparison here, which is that MySkills are often compared to another set of extremely popular engineering skills called superpowers.
chunk 6 · 3:28–4:04 · 125 words · 717 chars
3:28 And I'm really often asked the question, how do your skills compare to superpowers?
3:32 What's the difference between them?
3:34 To understand that, we need to understand the difference between user invoked and model invoked skills.
3:39 Anytime you have a skill, you can always invoke it manually.
3:43 So the skill sits on your file system.
3:45 The agent will just be able to pull up the skill and understand what's in there.
3:49 And you can always do that by communicating that to the agent.

Chapters

0:00 Introduction to the talk and the concept of "skill hell"
2:12 Overview of the skill checklist framework
3:16 Trigger: Choosing between user-invoked and model-invoked skills
7:29 Structure: Organizing steps and references
9:00 Making the skill.md file minimal
11:54 Steering: Using leading words to guide agent behavior
14:56 Increasing "leg work" per step
16:48 Pruning: Removing sediment, crud, and no-ops
19:06 Final summary of the checklist framework
19:55 Where to find the "writing great skills" resource