Skip to content

refactor(skills): pick skills by family, tags and traits alone - #1567

Draft
edwinyjlim wants to merge 2 commits into
dag/7-repackaged-skills-onlyfrom
dag/8-traits-and-tags-only
Draft

edwinyjlim wants to merge 2 commits into
dag/7-repackaged-skills-onlyfrom
dag/8-traits-and-tags-only

Conversation

@edwinyjlim

@edwinyjlim edwinyjlim commented Nov 17, 2026 •

Copy link
Copy Markdown
Member

Behaviour

Every skill is picked the same way: by its family, its tags and the project's traits.

  • Family. A menu entry's family is the key of the category the menu lists it under, with / as -. An entry under integration/install is in the family integration-install, and the 40 framework packs under integration are the family integration. The fetch sets it on every entry, and resolveSkill matches a family on it.
  • Tags. resolveSkill reads the framework and the default from an entry's tags, framework:<id> and default. The group, framework and default fields of a menu entry are no longer read.
  • Traits. A run's match keys are the project's traits, and nothing else. A run whose project has no framework trait resolves no family, so it has no framework pack.
  • One picker. install_skill resolves the name it is given through resolveSkill, so it accepts a menu id or a family. The spell book resolves a program's own skill and the framework's pack through resolveSkill too. With no framework it still hands over every skill whose family is the flow's name or is prefixed by it.

A bundle entry is still expanded into its variants and installed as before.

Verified

  • pnpm typecheck and pnpm lint (0 errors).
  • pnpm test: 292 files, 4,450 tests pass, 1 skipped.
  • On the recorded context-mill 1.66.0 menu, with three one-off scripts that are not part of this change:
    • Every entry's family equals its group, and every framework or default field has its tag: 502 entries, 200 of them bundle variants, with no exception. For every skill name and framework, the pick is the same as before: 38,544 comparisons, no difference.
    • The spell book hands over the same skills for every program, framework and session skill id: 1,008 cases, no difference.
    • Every program resolves the same framework pack and step skills with match keys from traits alone: 18 programs, 1,568 comparisons, no difference. No program's run label is a framework id.

Changes in behaviour to know about

  • install_skill now resolves a family. install_skill('integration') installs the framework's pack, and a family of bundle variants resolves to the variant for the run's framework.
  • A run with no detected framework no longer reports its program's label as the framework on skill events. It reports none.
  • The spell book no longer hands over a framework-less skill when the session's skill id names a family and no framework was detected. That needs wizard skill <family> on a project with no framework.

Rollout

This is the third of four stacked PRs that remove the old code paths. They merge together and ship in one release.

  • They wait until wizard-agent-dag and wizard-agent-dag-skills have both held at 100% with no stop rule tripped, no run that fetched its prompts, and no run that fell back to an old skill.
  • Both flags stay on in PostHog after the release. Released versions still read them, and a deleted flag reads as off.
  • Cloud moves to the release afterwards, with its context-mill pin at 1.66.0 or later.
  • This wizard reads no group, framework or default field, so context-mill can stop publishing them once older wizard versions have drained.

🤖 Generated with Claude Code

edwinyjlim and others added 2 commits November 16, 2026 17:11
A menu entry's family is the key of the category the menu lists it under,
with / as -. resolveSkill matches a family on that and reads framework and
default from tags only. The group, framework and default fields are no
longer read.

A run's match keys are the project's traits, with no override from the
run's input. install_skill and the spell book resolve a name through
resolveSkill, and the spell book reads a flow's families from the derived
family.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
@edwinyjlim
edwinyjlim added this pull request to stack #1569 November 17, 2026 01:22
@github-actions

Copy link
Copy Markdown

🧙 Wizard CI

Run the Wizard CI and test your changes against wizard-workbench example apps by replying with a GitHub comment using one of the following commands:

Test all apps:

  • /wizard-ci all

Test all apps in a directory:

  • /wizard-ci ai-observability
  • /wizard-ci basic-integration
  • /wizard-ci feature-flags
  • /wizard-ci mcp-analytics
  • /wizard-ci replay-vision
  • /wizard-ci revenue
  • /wizard-ci self-driving
  • /wizard-ci warehouse
  • /wizard-ci warehouse-seeded

Test an individual app:

  • /wizard-ci ai-observability/anthropic
  • /wizard-ci ai-observability/google-adk
  • /wizard-ci ai-observability/groq
Show more apps
  • /wizard-ci ai-observability/manual-capture
  • /wizard-ci ai-observability/openai
  • /wizard-ci ai-observability/openai-agents
  • /wizard-ci ai-observability/opentelemetry
  • /wizard-ci ai-observability/vercel-ai
  • /wizard-ci basic-integration/android
  • /wizard-ci basic-integration/angular
  • /wizard-ci basic-integration/astro
  • /wizard-ci basic-integration/django
  • /wizard-ci basic-integration/fastapi
  • /wizard-ci basic-integration/flask
  • /wizard-ci basic-integration/flutter
  • /wizard-ci basic-integration/javascript-node
  • /wizard-ci basic-integration/javascript-web
  • /wizard-ci basic-integration/laravel
  • /wizard-ci basic-integration/next-js
  • /wizard-ci basic-integration/nuxt
  • /wizard-ci basic-integration/python
  • /wizard-ci basic-integration/rails
  • /wizard-ci basic-integration/react-native
  • /wizard-ci basic-integration/react-router
  • /wizard-ci basic-integration/sveltekit
  • /wizard-ci basic-integration/swift
  • /wizard-ci basic-integration/tanstack-router
  • /wizard-ci basic-integration/tanstack-start
  • /wizard-ci basic-integration/vue
  • /wizard-ci feature-flags/django
  • /wizard-ci feature-flags/next-js
  • /wizard-ci mcp-analytics/custom-dispatcher
  • /wizard-ci mcp-analytics/go-sdk
  • /wizard-ci mcp-analytics/python-sdk
  • /wizard-ci mcp-analytics/ruby-sdk
  • /wizard-ci mcp-analytics/rust-sdk
  • /wizard-ci mcp-analytics/typescript-sdk
  • /wizard-ci replay-vision/javascript-node
  • /wizard-ci replay-vision/next-js
  • /wizard-ci replay-vision/react-vite
  • /wizard-ci revenue/stripe
  • /wizard-ci self-driving/astro
  • /wizard-ci self-driving/fastapi
  • /wizard-ci self-driving/nuxt
  • /wizard-ci self-driving/react-router
  • /wizard-ci self-driving/sveltekit
  • /wizard-ci warehouse/monorepo-env
  • /wizard-ci warehouse/multi-source-next
  • /wizard-ci warehouse/stripe-node
  • /wizard-ci warehouse/zero-source
  • /wizard-ci warehouse-seeded/next-stripe
  • /wizard-ci warehouse-seeded/next-stripe-declined

Test against a Context Mill branch:

  • /wizard-ci all context-mill:my-branch

Add context-mill:<branch> to any command above to pin the Context Mill branch. It defaults to main.

Results will be posted here when complete.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant