OpenAI Codex Adds Official Skills Support, Echoing Claude’s Open Agent Standard
OpenAI’s Codex now officially supports the open Agent Skills standard, introducing reusable instruction‑script‑resource bundles, progressive disclosure loading, explicit and implicit invocation methods, built‑in system skills, and cross‑tool compatibility including GitHub Copilot, signaling a move toward a unified AI agent ecosystem.
Codex Skills Overview
Codex Skills are reusable components that bundle Instructions , Scripts , and Resources . They enable Codex to perform specific workflows such as reading Linear tasks, converting PDFs to Markdown, or analyzing database schemas. Skills conform to the open Agent Skills standard.
Standard Skill Layout
SKILL.md– metadata (name, description) and the instruction set. scripts/ (optional) – executable scripts. references/ (optional) – documentation or reference material. assets/ (optional) – templates or other resources.
Execution Model
Codex uses a progressive disclosure mechanism: at startup it loads only the names and descriptions of all available Skills, conserving tokens. The full instruction set is fetched only when a Skill is invoked.
Skills can be invoked:
Explicit call – reference a Skill in the prompt with the $ prefix (e.g., $skill-name) or via the /skills command.
Implicit call – Codex automatically activates a Skill when the task description matches the Skill’s description.
Skill Scope Hierarchy
REPO level – $CWD/.codex/skills for project‑specific Skills.
REPO ROOT level – shared Skills for the whole repository.
USER level – ~/.codex/skills for personal tools usable across projects.
SYSTEM level – built‑in Skills shipped with Codex.
Built‑in System Skills
$skill-installer– installs Skills from GitHub. Example to add Linear support:
$skill-installer linear $skill-installer notion-spec-to-implementation– generates implementation code from a Notion API spec. $skill-creator – guides the user to create a SKILL.md file and optional scripts. $plan – performs research, produces a detailed execution plan, then carries out the steps.
GitHub Copilot Support
GitHub Copilot now supports Agent Skills. Skills placed in a repository’s .claude/skills directory are automatically recognized by Copilot, allowing the same Skill set to be used with Claude Code, Codex, and Copilot.
"If you’ve already set up skills for Claude Code in the .claude/skills directory in your repository, Copilot will pick them up automatically."
Current Copilot Skill coverage includes:
Copilot Coding Agent
Copilot CLI
Agent mode in VS Code Insiders (full VS Code support slated for January)
Remaining Inconsistencies
User‑level Skill directories differ: Claude Code uses ~/.claude/skills while Codex uses ~/.codex/skills. Installation conventions also vary across AI tools; the Agent Skills standard aims to unify these differences.
Signed-in readers can open the original source through BestHub's protected redirect.
This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactand we will review it promptly.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.
