Draft:Agent Skills

You can also browse Wikipedia:Featured articles and Wikipedia:Good articles to find examples of Wikipedia's best writing on topics similar to your proposed arti

Draft:Agent Skills
  • Comment: The sources are almost all primary, so for now, should exist under Claude or Anthropic. Potential COI with agent skills website. Either way not notable enough for standalone article until more (WP:THREE) secondary reliable (WP:RSP) sources cover the exact topic. Lacanic (talk) 03:12, 16 August 2026 (UTC)

Agent Skills
StatusPublished
Year started2025
First publishedOctober 16, 2025
OrganizationAnthropic
DomainAI agents
LicenseOpen specification
Websiteagentskills.io

Agent Skills is an open specification for packaging instructions, scripts, and reference files so that artificial intelligence agents can load them on demand. A skill is a folder built around a Markdown file named SKILL.md, which describes when the skill applies and how the agent should carry out the task.[1][2] Anthropic introduced the format for its Claude models in October 2025 and published it as an open standard on December 18, 2025.[1][3]

History

Anthropic released Skills for Claude on October 16, 2025, as a way to give the model task-specific instructions without keeping them in the context window at all times.[3][4] The developer Jesse Vincent had published a similar SKILL.md convention for structuring agent workflows about a week earlier.[4]

On December 18, 2025, Anthropic published the format as an open standard, releasing the specification and a software development kit at agentskills.io so that other AI platforms could adopt it.[1][5] The move followed Anthropic's donation of the Model Context Protocol to the Linux Foundation earlier that month.[1]

Design

A skill is a directory. The only required file is SKILL.md, which starts with a YAML header giving the skill's name and a short description, followed by Markdown instructions. The folder can also hold scripts, templates, and reference documents.[2][4]

The format relies on what the specification calls progressive disclosure. An agent first reads only the name and description of each installed skill, which costs a few dozen tokens. It loads the full instructions, and any bundled files, only when a user request matches the skill.[6][4] Because a skill is plain files rather than code tied to one vendor, the same skill can run on any agent that supports the format.[5]

Adoption

Microsoft, OpenAI, Atlassian, Figma, Cursor, and GitHub adopted the standard when it was opened in December 2025.[6][5] Anthropic also launched a skills directory with partner-built skills from Canva, Stripe, Notion, and Zapier.[1] As of August 2026, the project's client showcase listed more than forty agent products that support the format, including Google's Gemini CLI, GitHub Copilot, Visual Studio Code, and OpenAI's Codex.[7]

According to the Agent Economy Census published by the directory Skillselion, 79,827 skill listings had been published across public registries by August 2026, with about 158 million cumulative installs, and the most-installed 1% of skills accounted for 77% of all installations.[8] A 2026 case study of a skills deployment at Yahoo reported that a survey of 67 engineers found an average time saving of 2.6 hours per week.[9]

Reception

The programmer Simon Willison called the October 2025 release "maybe a bigger deal than MCP", arguing that the low cost of writing and sharing a skill would drive faster adoption than the Model Context Protocol saw.[4] Holger Mueller of Constellation Research was more cautious, noting that single vendors rarely manage to set industry standards on their own and that the wide endorsement of MCP did not guarantee the same outcome for Agent Skills.[1]

See also

References

  1. ^ a b c d e f Deutscher, Maria (December 18, 2025). "Anthropic makes agent Skills an open standard". SiliconANGLE. Retrieved August 12, 2026.
  2. ^ a b "Agent Skills specification". Retrieved August 12, 2026.
  3. ^ a b "Introducing Agent Skills". Anthropic. October 16, 2025. Retrieved August 12, 2026.
  4. ^ a b c d e Willison, Simon (October 16, 2025). "Claude Skills are awesome, maybe a bigger deal than MCP". Retrieved August 12, 2026.
  5. ^ a b c "Anthropic launches enterprise 'Agent Skills' and opens the standard, challenging OpenAI in workplace AI". VentureBeat. December 18, 2025. Retrieved August 12, 2026.
  6. ^ a b "Anthropic Opens Agent Skills Standard, Continuing Its Pattern of Building Industry Infrastructure". Unite.AI. December 19, 2025. Retrieved August 12, 2026.
  7. ^ "Agent Skills — Client Showcase". agentskills.io. Retrieved August 12, 2026.
  8. ^ "Skillselion Agent Economy Census, August 2026". Skillselion. August 2026. Retrieved August 12, 2026.
  9. ^ Bakal, Gal (March 2026). "Knowledge Activation: AI Skills as the Institutional Knowledge Primitive for Agentic Software Development". arXiv:2603.14805 [cs.SE].

Content Disclaimer

Informasi ini disarikan dari Wikipedia dan disajikan kembali untuk tujuan edukasi. Konten tersedia di bawah lisensi CC BY-SA 3.0. Kami tidak bertanggung jawab atas ketidakakuratan data yang bersumber dari kontribusi publik tersebut.

  1. The information displayed on this website is sourced in part or in whole from Wikipedia and has been adapted for the purpose of restating it. We strive to provide accurate and relevant information, however:
  2. There is no guarantee of absolute accuracy. Wikipedia is an open, collaborative project that can be edited by anyone, so information is subject to change.
  3. It is not intended to constitute professional advice. The content displayed is for informational and educational purposes only. For important decisions (e.g., medical, legal, or financial), please consult a professional.
  4. Content copyright. Wikipedia is licensed under the Creative Commons Attribution-ShareAlike License (CC BY-SA). This means that content may be reused with appropriate attribution and shared under a similar license.
  5. Responsible use. Any risk arising from the use of information from this website is entirely the responsibility of the user.