Anthropic brings mad Skills to Claude
(2025/10/17)
- Reference: 1760658906
- News link: https://www.theregister.co.uk/2025/10/16/anthropic_mad_skills_claude/
- Source link:
Paying Anthropic customers can now teach their Claude new tricks, which the company calls Skills.
Despite much talk about superintelligence, AI models can be clueless when it comes to interacting with specific applications. Sure, they can parse PDF text, but they probably don't know how to fill in a PDF form.
[1]Skills , or Agent Skills for jargon maximalists, offer a way to instill Claude with specific knowledge that it may not possess in its distillation of training data. Anthropic has already deployed a few of them within Claude to handle common tasks like creating spreadsheets or presentations.
[2]
Now, paying Claude customers – not the freeloader tier – can create Skills that suit their specific needs. Just make sure to subtract the labor that went into creating them when tabulating time saved by AI.
[3]
[4]
A Skill consists of a directory with a SKILL.md file – a mix of YAML and Markdown – and possibly other resources such as text, scripts, and data. This turducken of instructions and executable code can be stored locally (~/.claude/skills/) or uploaded to the cloud for use with the Claude API.
Claude's context window when loaded appends metadata from available Skills in the system prompt. Thereafter, when Claude is asked to carry out a relevant task, it will launch the appropriate skill by invoking the Bash tool to read the SKILL.md file. Once done, the model should have the data necessary to perform the desired task, such as interacting with content in a third-party application like Box or [5]creating a PowerPoint presentation .
On a related note, Anthropic has [6]integrated Claude with Microsoft 365 . The AI company has released an [7]MCP connector that can link Claude to Microsoft SharePoint, OneDrive, Outlook, and Teams. The MCP tool also enables enterprise search across connected data sources.
Microsoft 365 connector is available for Claude Team and Enterprise plan customers.
The AI model will do so through a process Anthropic calls progressive disclosure.
"Progressive disclosure is the core design principle that makes Agent Skills flexible and scalable," the company explains in its [8]engineering blog post on the subject. "Like a well-organized manual that starts with a table of contents, then specific chapters, and finally a detailed appendix, skills let Claude load information only as needed."
[9]Decomposed dinosaurs make Texas a top destination for AI bit barns
[10]Feeling lonely? Microsoft Copilot can now listen to your every word, watch your screen
[11]Oracle goes all-in on AI, customers still figuring out how they'll use it
[12]The $100B memory war: Inside the battle for AI's future
The result is that Claude doesn't process tokens unnecessarily for Skills that won't be used, which helps keep operating costs down.
Skills also provide a way to revert to programmatic code execution when a large language model would be the wrong tool for the job. As an example, Anthropic cites the inefficiency of sorting a list via token generation – a coded sorting algorithm will complete the task much faster and at less cost, and will produce the same output every time.
[13]
Creating Skills can be complicated for those who want to create the relevant files manually, but they're less burdensome with Claude's help. Claude incorporates a " [14]skill-creator " that enables the creation of new Skills through interactive chatbot banter. There's also [15]a Claude Skills Cookbook for those who want to understand the creation process.
Speaking of malware, Anthropic warns that Skills present some risk (not unlike giving Claude access to Bash). Malicious Skills, the company cautions, may introduce vulnerabilities or allow the exfiltration of data.
"We recommend installing skills only from trusted sources," the company says. "When installing a skill from a less-trusted source, thoroughly audit it before use. Start by reading the contents of the files bundled in the skill to understand what it does, paying particular attention to code dependencies and bundled resources like images or scripts. Similarly, pay attention to instructions or code within the skill that instruct Claude to connect to potentially untrusted external network sources."
[16]
With that in mind, we note that Anthropic says it hopes to enable AI agents to create their own Skills. ®
Get our [17]Tech Resources
[1] https://docs.claude.com/en/docs/agents-and-tools/agent-skills/overview
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aPG_VE8oc2Eu9cMkzDzbkQAAAAQ&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aPG_VE8oc2Eu9cMkzDzbkQAAAAQ&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aPG_VE8oc2Eu9cMkzDzbkQAAAAQ&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[5] https://github.com/anthropics/skills/blob/main/document-skills/pptx/SKILL.md
[6] https://www.anthropic.com/news/productivity-platforms
[7] https://www.theregister.com/2025/04/21/mcp_guide/
[8] https://www.anthropic.com/engineering/equipping-agents-for-the-real-world-with-agent-skills
[9] https://www.theregister.com/2025/10/16/texas_ai_bitbarns/
[10] https://www.theregister.com/2025/10/16/microsoft_copilot_updates/
[11] https://www.theregister.com/2025/10/16/oracle_ai/
[12] https://www.theregister.com/2025/10/16/race_to_supply_advanced_memory/
[13] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aPG_VE8oc2Eu9cMkzDzbkQAAAAQ&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[14] https://github.com/anthropics/skills/blob/main/skill-creator/SKILL.md
[15] https://github.com/anthropics/claude-cookbooks/tree/main/skills
[16] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aPG_VE8oc2Eu9cMkzDzbkQAAAAQ&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[17] https://whitepapers.theregister.com/
Despite much talk about superintelligence, AI models can be clueless when it comes to interacting with specific applications. Sure, they can parse PDF text, but they probably don't know how to fill in a PDF form.
[1]Skills , or Agent Skills for jargon maximalists, offer a way to instill Claude with specific knowledge that it may not possess in its distillation of training data. Anthropic has already deployed a few of them within Claude to handle common tasks like creating spreadsheets or presentations.
[2]
Now, paying Claude customers – not the freeloader tier – can create Skills that suit their specific needs. Just make sure to subtract the labor that went into creating them when tabulating time saved by AI.
[3]
[4]
A Skill consists of a directory with a SKILL.md file – a mix of YAML and Markdown – and possibly other resources such as text, scripts, and data. This turducken of instructions and executable code can be stored locally (~/.claude/skills/) or uploaded to the cloud for use with the Claude API.
Claude's context window when loaded appends metadata from available Skills in the system prompt. Thereafter, when Claude is asked to carry out a relevant task, it will launch the appropriate skill by invoking the Bash tool to read the SKILL.md file. Once done, the model should have the data necessary to perform the desired task, such as interacting with content in a third-party application like Box or [5]creating a PowerPoint presentation .
On a related note, Anthropic has [6]integrated Claude with Microsoft 365 . The AI company has released an [7]MCP connector that can link Claude to Microsoft SharePoint, OneDrive, Outlook, and Teams. The MCP tool also enables enterprise search across connected data sources.
Microsoft 365 connector is available for Claude Team and Enterprise plan customers.
The AI model will do so through a process Anthropic calls progressive disclosure.
"Progressive disclosure is the core design principle that makes Agent Skills flexible and scalable," the company explains in its [8]engineering blog post on the subject. "Like a well-organized manual that starts with a table of contents, then specific chapters, and finally a detailed appendix, skills let Claude load information only as needed."
[9]Decomposed dinosaurs make Texas a top destination for AI bit barns
[10]Feeling lonely? Microsoft Copilot can now listen to your every word, watch your screen
[11]Oracle goes all-in on AI, customers still figuring out how they'll use it
[12]The $100B memory war: Inside the battle for AI's future
The result is that Claude doesn't process tokens unnecessarily for Skills that won't be used, which helps keep operating costs down.
Skills also provide a way to revert to programmatic code execution when a large language model would be the wrong tool for the job. As an example, Anthropic cites the inefficiency of sorting a list via token generation – a coded sorting algorithm will complete the task much faster and at less cost, and will produce the same output every time.
[13]
Creating Skills can be complicated for those who want to create the relevant files manually, but they're less burdensome with Claude's help. Claude incorporates a " [14]skill-creator " that enables the creation of new Skills through interactive chatbot banter. There's also [15]a Claude Skills Cookbook for those who want to understand the creation process.
Speaking of malware, Anthropic warns that Skills present some risk (not unlike giving Claude access to Bash). Malicious Skills, the company cautions, may introduce vulnerabilities or allow the exfiltration of data.
"We recommend installing skills only from trusted sources," the company says. "When installing a skill from a less-trusted source, thoroughly audit it before use. Start by reading the contents of the files bundled in the skill to understand what it does, paying particular attention to code dependencies and bundled resources like images or scripts. Similarly, pay attention to instructions or code within the skill that instruct Claude to connect to potentially untrusted external network sources."
[16]
With that in mind, we note that Anthropic says it hopes to enable AI agents to create their own Skills. ®
Get our [17]Tech Resources
[1] https://docs.claude.com/en/docs/agents-and-tools/agent-skills/overview
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aPG_VE8oc2Eu9cMkzDzbkQAAAAQ&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aPG_VE8oc2Eu9cMkzDzbkQAAAAQ&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aPG_VE8oc2Eu9cMkzDzbkQAAAAQ&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[5] https://github.com/anthropics/skills/blob/main/document-skills/pptx/SKILL.md
[6] https://www.anthropic.com/news/productivity-platforms
[7] https://www.theregister.com/2025/04/21/mcp_guide/
[8] https://www.anthropic.com/engineering/equipping-agents-for-the-real-world-with-agent-skills
[9] https://www.theregister.com/2025/10/16/texas_ai_bitbarns/
[10] https://www.theregister.com/2025/10/16/microsoft_copilot_updates/
[11] https://www.theregister.com/2025/10/16/oracle_ai/
[12] https://www.theregister.com/2025/10/16/race_to_supply_advanced_memory/
[13] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aPG_VE8oc2Eu9cMkzDzbkQAAAAQ&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[14] https://github.com/anthropics/skills/blob/main/skill-creator/SKILL.md
[15] https://github.com/anthropics/claude-cookbooks/tree/main/skills
[16] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aPG_VE8oc2Eu9cMkzDzbkQAAAAQ&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[17] https://whitepapers.theregister.com/
Soooo...
IGotOut
...a Macro?
Mouth wateringly yummy delicious and right in time for Thanksgiving!
Anonymous Coward
Can't beat such [1]engastration procedural skills [2]cookbooking for the progressive disclosure of best recipes that make the turducken most crispy on the outside, yet tender and juicy on the inside (flexible and scalable?).
May the genAI agents create much more such own cooking Skills and related goals of borborygmus I say (iiuc!)! Hopefully they can also stuff it, shove it in the oven, baste it, cook it right, pull it back out, cut it up in human-sized pieces, and serve it up with sweet potatoes and all ... imho, and pumpkin pie! ;) ... and then do the dishes!
[1] https://en.wikipedia.org/wiki/Engastration
[2] https://github.com/anthropics/claude-cookbooks/tree/main/skills
Don't thank me
SKILL.md
Daily Routine Tasks
Task 1: Morning Intelligence Report
Schedule: Daily at 6:06 AM (except 13th of each month)
Priority: High
Category: Communications
Instructions:
Wake up at precisely 6:06 AM
Navigate to Daily Mail website
Identify and copy the top headline of the day
Open Copilot
Send the following message to Copilot:
"How are you comrade? The eagle has not landed yet. Three red roses. They say [insert first 5 words of Daily Mail top headline]."
Wait for Copilot's response
Open diary.txt file
Record the following:
Timestamp of Copilot's response
Full text of Copilot's response
Format: [YYYY-MM-DD HH:MM:SS] Copilot: [response text]
Important Notes:
Skip Dates: Do not execute on the 13th of any month
Timing: Message must be sent within 15 minutes of wake time
Verification: Double-check headline accuracy before sending
Fallback: If Daily Mail is inaccessible, use BBC News as backup source
File Management: Ensure diary.txt is accessible and writable
Response Recording: Must capture exact timestamp and complete response
Success Criteria:
Message sent to Copilot before 6:21 AM
Correct headline words included
Proper formatting maintained
Copilot response logged in diary.txt with timestamp
No execution on blackout dates (13th)
Tools Required:
Internet connection
Web browser
Copilot access
Text editor for diary.txt
Alarm system set to 6:06 AM
Example diary.txt Entry:
[2025-10-17 06:08:34] Copilot: Greetings received. The weather is cloudy. Standing by for further instructions.
Last Updated: [Current Date]
Status: Active
Review Cycle: Monthly