Skip to content

Release v7.14.3

Choose a tag to compare

View all tags
@github-actions github-actions released this 08 Oct 22:17
v7.14.3
d01946b
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Summary

The skills now follow Anthropic's skill authoring best practices and the Agent Skills specification: every SKILL.md stays within the instruction budget, long reference files open with a contents list, and every reference is linked directly from SKILL.md with the situation that calls for it. No guidance was removed; sections that only some tasks need moved verbatim into reference files.

Changed

  • Ten skills that exceeded 500 lines or about 5,000 tokens now keep their decision rules, traps and version pins in SKILL.md and move secondary detail into reference files: the .NET, TypeScript, Elixir and Kotlin Chronicle clients, reactors, Arc commands in TypeScript, and the Screenplay capture, command-surface, event-modeling and specifications skills (#542)
  • Reference files longer than 100 lines open with a contents list, so an agent that previews the top of a file still sees everything it covers (#542)
  • Links to reference files say when to read them, and traps an agent must know before acting now sit in SKILL.md ahead of the material that depends on them (#542)
  • Capability limits are stated against the version each skill was verified on instead of "today" or "yet" (#542)
  • Screenplay skills name the screenplay MCP server together with each tool they call (#542)
  • The read-model skill is pinned to Arc 22.16.0 and the performance review to Components 4.26.2, the versions their APIs were verified against (#542)
  • Six skill descriptions no longer contain generic type brackets such as IProjectionFor<T>, which the Claude Skills API rejects (#542)

Fixed

  • The Screenplay model review skill routes Chronicle CLI questions to cratis-chronicle-cli-operations instead of a skill that does not exist (#542)
  • The Lens skill routes executing a command from backend code to the right skill, and the React page dialog reference points to the step that registers confirmation and busy-indicator dialogs (#542)
  • The Screenplay specifications skill states that the reference runner executes clock, trigger and capture actions from ESM v6 (#542)
  • The identity skill states in SKILL.md that a resolved tenant is never an authorization decision (#542)
  • The release-notes skill reads a repaired release body back before reporting the repair as done (#542)
Assets 2
Loading