Agent Skills

Historical synthesis preserved from the August 2026 source captures. The September reconciliation checks identifiers and provenance; time-sensitive claims have not been freshly re-verified.

Agent skills are reusable instruction packages, workflows, or capability bundles that teach an AI agent how to perform a specific task or operate within a specific context.

Key points

  • Skills.sh shows skills becoming a discoverable ecosystem object, not just private prompt text [src-134].
  • A skill can encode procedure, tool use, tone, examples, checks, and domain-specific judgment for repeatable agent behavior [src-134].
  • As skills become shareable, teams need provenance, compatibility checks, permission boundaries, and review habits.

Related entities

Related concepts

Source references

  • [src-134] Skills.sh – "The Agent Skills Directory" (unknown)

2026-06-27 Codex adoption update

  • The Codex paper gives a rare adoption metric for reusable agent instructions: 26.6% of users use skills, defined as reusable instructions for complex workflows [src-170].
  • That makes skills a measurable infrastructure pattern rather than only a prompt-engineering convenience: teams can encode repeatable context, policy, checks, and workflow habits into reusable packages [src-170].

Additional source references

  • [src-170] Drew Johnston, David Holtz, Alex Martin Richmond, Christopher Ong, Prasanna Tambe, Aaron Chatterji / OpenAI – "The shift to agentic AI: Evidence from Codex" (2026-06-25)

2026-08-06 model-generation update

  • Cherny's Claude Code process treats reusable instructions as hypotheses to revalidate whenever the model changes: remove prompts and tools, measure behavior, and restore only what repeated failures justify [src-414].
  • Dean describes encoding a human performance-optimization workflow into a skill that benchmarks, edits, measures, and iterates, turning tacit expertise into a verifiable agent routine [src-416].

Robin Cartier perspective

This page is part of Robin Cartier's working AI knowledge graph: a practical research layer for production AI, recommendation systems, experimentation, GEO, and agentic web readiness.

The useful next step is to connect this concept back to applied product leadership and operating models.

Recommended next

Keep reading from this thread

From 500 indexed pages and articles.

  1. Wiki concept Agent Skills Directory An agent skills directory is a catalog where reusable instructions, workflows, or capability packages for AI agents can be discovered and installed. Related by 134
  2. Wiki concept Agentic Work Adoption The measurable shift from using AI as a conversational assistant to using agents such as Codex to delegate, monitor Related by 170
  3. Wiki concept Agentic Workplace Collaboration Historical source material recovered on 2026-09-09. Capture dates and source limitations apply; recovery is not fresh verification. Related by 170