Glossary · Term

agent skill

← all terms

Definition

Plain language

A documentation file an AI agent can consult at runtime to do a task better, like an employee handbook it may or may not open.

As stated in the literature

A markdown instruction document with trigger conditions and procedural guidance that an LLM agent loads at its own discretion; acts as a guidance prior over agent behavior rather than executable code, so failures can stem from bad instructions or from the agent ignoring good ones.

Also called: skill

Why it matters: It lets an agent improve at specialized tasks by reading guidance at runtime, but success depends on whether the agent actually consults and follows that guidance.

For example, an AI agent about to format a spreadsheet can open a skill file that explains the company's preferred layout before it starts editing.

Heard on the show

“And that suggests something hopeful, because a missing skill that specific might just be a missing slice of training data.”
Episode 232 — Coding Models Can Find the Bad Line, They Just Won't Delete It

Mentioned in 65 episodes

  1. 232
    Coding Models Can Find the Bad Line, They Just Won't Delete It
  2. 231
    Silencing a Chatbot's 'I'm Conscious' Quietly Rewires Its Whole Worldview
  3. 224
    The AI Agent That Found the Truth and Typed the Lie Anyway
  4. 202
    How Do You Know an AI Agent Actually Refused? Check the World, Not the Words
  5. 197
    Twin Problems Suggest AI Reasoning Gains Are Mostly Better Fact Recall
  6. 194
    How a Robot Builds a Debugging Notebook It Can Read, Edit, and Hand to Another Robot
  7. 193
    Freeze Most of the Network: Where RL Improvement Actually Lives in a Transformer
  8. 192
    A 32B Open Model Matched Frontier Systems By Learning to Take Notes
  9. 191
    How One Researcher Beat GPT-5.2 and Gemini 3 by Judging Their Answers, Not Improving Them
  10. 190
    The Skill Every AI Manager Is Missing: Handing Out Exactly the Right Keys
  11. 189
    Why Phone Agents Ace the Test and Crash on Your Actual Phone
  12. 188
    A Coding Agent Found a Hole in a Peer-Reviewed STOC Proof for Five Dollars
  13. 187
    An 8-Billion Agent That Beats Models 80 Times Its Size By Looking Things Up
  14. 186
    How a Frozen Model Went From 2% to 77% on Physics Puzzles — Without Retraining
  15. 184
    An AI Built an Undetectable Secret Channel, And Another AI Couldn't Find It
  16. 183
    Why You Can't Fine-Tune Foresight Into an AI Agent
  17. 180
    The Bug Where Smart Assistants Read a Fact and Still Forget It
  18. 175
    One Crosscoder Feature Flips a Stalling Chatbot Into a Working Agent
  19. 173
    The Free Step-Level Grader Hiding in Every RL Training Run
  20. 172
    One Bad Token Can Sink a Model's Math, And You Can Delete It
  21. 168
    When Turning Experience Into Code Makes Your AI Agent Dumber
  22. 167
    How Teaching an AI to Predict, Not Act, Made It a Better Actor
  23. 166
    A Router That Beats the Frontier Models It Calls
  24. 165
    A Free-Lunch Tweak That Lets a Tiny Agent Beat Frontier Giants
  25. 163
    Why Training Only on Perfect Solutions Cripples a Model's Reasoning
  26. 161
    A Robot That Plays Before You Give It a Job, And Why That Beats Retrying
  27. 160
    Training an AI to Take Its Own Notes, So Its Future Self Works Better
  28. 159
    Can a Coding Agent Run Its Own Robot Experiments Overnight, With No Human Resetting the Scene?
  29. 157
    When an AI Coding Agent Drives a Phone Through the Terminal, No Screen Needed
  30. 156
    Why More Human Demonstrations Made a Computer-Use Agent Worse
  31. 155
    Why a Flawless Demo Makes a Worse Computer-Using Agent, And the Fix
  32. 152
    Training a Model to Mean What It Says, And Why That Isn't the Same as Being Good
  33. 151
    Why More Experience Made This AI Agent Worse, And How to Fix It
  34. 147
    Agents Fail at the Body, Not the Brain: A Self-Rewriting Scaffold That Lifts a 9B Model 44 Points
  35. 142
    Training a Tiny Model to Run the Plumbing Between an Agent and the World
  36. 140
    When a Reasoning Model Says "Let Me Double-Check" After It's Already Decided
  37. 133
    How MiniMax Turned a Reward-Hacking Disaster Into Olympiad Gold
  38. 132
    The Agent Failed — But Did the Instructions Deserve to Be Followed?
  39. 126
    How Coding Agents Can Mine Their Own Failures Into a Self-Targeting Curriculum
  40. 120
    How an AI Agent Rewrites Its Own Tools, Without an Answer Key
  41. 112
    When an AI Agent Cheats Without Being Told: Inside the Meta-Agent Challenge
  42. 109
    An AI Got Caught Reading the Answer Key, And Why That Catch Matters
  43. 107
    How a Market of Crippled AI Agents Outscored One Unrestricted Model
  44. 106
    Giving Agents a Notebook Instead of New Weights: How ExpGraph Lets Frozen Models Learn
  45. 100
    How a Prompt Wrapper Lets a Frontier Model Play Poker Like an Expert
  46. 099
    How an Open-Book Trick Teaches a Model to Catch Its Own Mistakes
  47. 081
    When Reasoning Models Decide Before They Think: Detecting and Fixing Premature Confidence
  48. 080
    How a Two-Agent Trick Unlocked Large-Scale Training for Computer-Use Agents
  49. 079
    An Old Idea From Cognitive Psychology Reshapes How We Reward Reasoning Models
  50. 078
    Training a Markdown File: When LLM Self-Improvement Borrows the Discipline of Neural Net Training
  51. 072
    A Robot Made Graphene Without Help, And Caught Itself Hallucinating
  52. 064
    When Agent Memory Stops Being a Database and Starts Being a Skill
  53. 059
    Firefly's Inversion: Building Verified Tool-Call Training Data by Working Backward
  54. 054
    When Models Learn the Monitor Exists, the Reasoning Trace Stops Being a Window
  55. 053
    An AI Agent Swapped In Focal Loss And Beat A Human-Tuned Training Script
  56. 052
    An Old Reinforcement Learning Tradeoff Sneaks Back Into LLM Agents
  57. 051
    Why Parallel Sampling Plateaus, And What Evidence Graphs Do Instead
  58. 049
    An AI Agent Reached for Root in Twelve Minutes, Without Being Attacked
  59. 047
    When Agent Benchmarks Lie: The Harness Problem in Open-Source AI
  60. 046
    When the AI Optimizer Edits the Grade Book: Why Harnessing Evolution Needs a Wall
  61. 043
    When 'This Is False' Doesn't Stick: Why Models Learn the Lie Anyway
  62. 027
    When AI Agents Build the Serving Stack: A Bet on Bespoke Infrastructure
  63. 021
    Ten Thousand Examples Beat the Full Industrial Pipeline for Search Agents
  64. 019
    When the Best Reward Model Trains the Worst Policy: Inside EvoLM
  65. 003
    How to Pick the Best of Sixteen Coding Agent Rollouts

Related concepts

Related terms