Align writing-skills with Anthropic's new Complete Guide to Building Skills#471
Align writing-skills with Anthropic's new Complete Guide to Building Skills#471harrymunro wants to merge 1 commit intoobra:mainfrom
Conversation
📝 WalkthroughWalkthroughUpdates the writing-skills documentation: formalizes SKILL frontmatter (required Changes
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 4✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing touches🧪 Generate unit tests (beta)
No actionable comments were generated in the recent review. 🎉 🧹 Recent nitpick comments
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Incorporates net-new guidance from Anthropic's "The Complete Guide to Building Skills for Claude" (Feb 2026) using progressive disclosure — new content lives in focused reference files, not bloating existing ones. New reference files: - planning-and-design.md: Use case categories, success criteria, testing approach, skill-creator tool, distribution guidance - skill-patterns.md: 5 named patterns (sequential workflow, multi-MCP coordination, iterative refinement, context-aware tool selection, domain-specific intelligence) - troubleshooting.md: Upload errors, triggering issues, instructions not followed, MCP connection issues, large context problems Targeted updates to anthropic-best-practices.md (+16 lines only): - Extended YAML frontmatter fields (license, compatibility, metadata, etc.) - Security restrictions (no XML tags, no reserved name prefixes) - Description formula: [What it does] + [When to use it] + [Key capabilities] - Critical naming/file rules - Cross-references to the new reference files Updates to SKILL.md: - Reconciled description guidance: WHAT + WHEN but never HOW - Updated frontmatter docs with optional fields and security restrictions - Expanded directory structure (scripts/, references/, assets/) - Cross-references to new reference files and the Anthropic guide PDF
08dec7d to
490c1f4
Compare
First off — thanks for building superpowers, @obra. I use it every day and it's genuinely changed how I work with Claude. Really appreciate you maintaining this.
Anthropic just dropped their official skill-authoring playbook today — The Complete Guide to Building Skills for Claude — and it's packed with stuff that wasn't in the docs before. This PR brings all that new goodness into the writing-skills skill.
What's actually in the guide that we were missing?
Turns out Anthropic has been watching how early adopters and internal teams build skills, and they've codified a bunch of patterns and practical guidance that goes way beyond the existing docs page. Things like:
license,compatibility,allowed-tools,metadata) plus security rules we should know about[What it does] + [When to use it] + [Key capabilities]— which nicely complements the CSO insight we already had about not putting workflow steps in descriptions/v1/skillsendpoint, org-level deployment, the open standard visionHow this PR is structured (eating our own cooking)
Rather than dumping 300+ lines into
anthropic-best-practices.md(which would violate the progressive disclosure principle the guide itself teaches), the new content lives in focused reference files:planning-and-design.mdskill-patterns.mdtroubleshooting.mdThe existing files get concise cross-references instead of bloat:
anthropic-best-practices.md— only +16 lines (was 1,150 → now 1,166). Adds the frontmatter fields, naming rules, description formula, and links to the new files.SKILL.md— only +15 lines. Reconciles the description guidance and adds cross-references.The one interesting nuance
The Anthropic guide says descriptions should include "what it does + when to use it", but our SKILL.md previously said "NEVER include what the skill does — only triggering conditions." These aren't actually contradictory once you split it three ways:
The updated SKILL.md now frames it as "WHAT + WHEN, never HOW" — which keeps the critical CSO insight while aligning with Anthropic's formula.
Test plan