wwdc-knowledge
Skill2 starsChanged 40 days agoMITView on GitHub
Brings Apple's WWDC guidance into development — session transcripts, Apple's own code snippets, and the documentation each session links, as primary sources across 1638 sessions from 2014 to 2026 plus Tech Talks and Meet with Apple. Use when writing or reviewing code for an Apple platform and Apple has a recommended approach (SwiftUI, UIKit, Swift concurrency, Foundation Models, SwiftData, Metal, StoreKit); when adopting a framework or choosing between APIs; when an availability or deprecation claim needs pinning to a version; when the user names a WWDC session, a year, or asks what changed in a framework; or when another skill needs Apple's stated guidance rather than recall.
In this repository28 other files it gives its agents.
- broSkill
- codebase-designSkill
- diagnosing-bugsSkill
- domain-modelingSkill
- grill-with-docsSkill
- implementSkill
- implement-specSkill
- improve-codebase-architectureSkill
- iphone-duo-adaptSkill
- openai-docsSkill
- product-thinkingSkill
- prSkill
- researchSkill
- retroSkill
- setup-matt-pocock-skillsSkill
- swift-concurrency-expertSkill
- swift-concurrencySkill
- swiftui-expert-skillSkill
- swiftui-liquid-glassSkill
- swiftui-performance-auditSkill
- swiftui-proSkill
- swiftui-ui-patternsSkill
- swiftui-view-refactorSkill
- tddSkill
- to-specSkill
- to-ticketsSkill
- typesafe-aiSkill
- wayfinderSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 11
Every claim you return carries a timecode: wwdc2025-286 [4:01].
- Very specific: line 44
Prefer --at over --full: it returns the exact sentences to quote, and --full can run past 20 KB.
- Distinctive: line 58
When the output is code rather than prose, the citation goes in the message beside it, naming the session that recommends the approach.
- Very specific: line 68
When they conflict, say which year each came from.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
WWDC Knowledge
Apple's engineers explained these frameworks on stage, and said how they intend them to be used. Those talks are the primary source for both, and this skill reads them locally.
Every claim you return carries a timecode: wwdc2025-286 [4:01]. The timecode is what separates
a sourced answer from recall wearing a citation's clothes.
wwdc.py below means python3 <this skill's directory>/scripts/wwdc.py. Use the absolute path.
Steps
Find the sessions.
wwdc.py search "guided generation" wwdc.py search "scroll performance" --topic swiftui --event wwdc2026 wwdc.py search --regex '@Generable\b' --excerpts 1 wwdc.py search --topic ai --event wwdc2026 # survey a slice, no queryQuery terms are ANDed within a single transcript paragraph, so two or three specific words beat a sentence. Use
--regexfor exact API spellings that spoken prose mangles. Filter vocabulary for--topic,--event, and--platformcomes fromwwdc.py info.Output gives you a session id, a timecode per excerpt, and the dataset path. Carry the id forward; it is what
showtakes.Read what the excerpt promised. The search excerpt is a pointer, not the answer.
wwdc.py show wwdc2025-286 # summary, counts, what's available wwdc.py show wwdc2025-286 --at 4:01 --pad 60 # sentence-level transcript around a timecode wwdc.py show wwdc2025-286 --code # Apple's snippets, each with its timecode wwdc.py show wwdc2025-286 --resources # linked docs, with markdown URLs wwdc.py show wwdc2025-286 --full # whole transcriptPrefer
--atover--full: it returns the exact sentences to quote, and--fullcan run past 20 KB. Reach for--codewhenever the answer involves writing Swift, because Apple's own snippet beats one you compose.- Follow the API into Apple's documentation. Each resource carries a
sosumiURL, which renders adeveloper.apple.compage as Markdown. Fetch it when the session names a symbol whose signature, parameters, or availability you need. Any Apple docs URL converts the same way: swap the host forsosumi.ai. Answer in timecodes. Done when every API name, signature, availability claim, and line of code you produce traces to a transcript segment, a code snippet, or a doc page you opened this turn, each carrying its session id and timecode. A statement you cannot attribute that way is one you still need to look up, or one you present plainly as your own inference.
When the output is code rather than prose, the citation goes in the message beside it, naming the session that recommends the approach.
Quote the sentences that carry the point and link the session. These transcripts are Apple's copyrighted material, so short cited excerpts are the shape to use.
Dating a claim
The corpus spans twelve years, so two sessions can both be right about different years. Prefer the
newest session covering the API, and read an older talk as a record of how the API stood that year.
When they conflict, say which year each came from. search sorts newer events first, so the top hit
is usually the current one.
Dataset
Sessions live in a clone of github.com/guitaripod/wwdc-sessions, at $WWDC_SESSIONS_DIR or
~/.local/share/wwdc-sessions. wwdc.py prints the exact fix if it is missing.
- Absent — ask the user before running
wwdc.py setup; it downloads about 175 MB. - Stale —
wwdc.py refreshfast-forwards it.searchfooters andwwdc.py infoshow the build date. Apple publishes in bursts around June, so a refresh matters most right after a WWDC. - 87 sessions have no transcript and are flagged
[no transcript]; their metadata still resolves.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

