Repository navigation
refactor(orchestrator): read agent prompts from the wizard only - #1565
Conversation
The orchestrator builds its registry from the prompts a program bundles and picks skills with the shared resolver and the project's match keys. The wizard-agent-dag flag, the remote prompt fetch and the old menu lookup are removed. A program that bundles no prompts, or whose bundle fails to load, fails the run with PHW_AGENT_ORCHESTRATOR_NO_PLANNER. A run pinned with --context-mill-version reads the bundled prompts too; the pin selects skills only. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
🧙 Wizard CIRun 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:
Test all apps in a directory:
Test an individual app:
Show more apps
Test against a Context Mill branch:
Add Results will be posted here when complete. |
johncwaters
left a comment
There was a problem hiding this comment.
Verdict: REQUEST CHANGES
See inline comments.
Automated review by Glimmervoid. Not written by a human
gewenyu99
left a comment
There was a problem hiding this comment.
🥸 Reviewed by NotVincent, totally not Vincent. The real Vincent will review it separately. Probably slop, please disregard.
|
✅ Security review complete — No findings above the confidence threshold, in 6 min. Add the |
…and every framework resolves its skills A program that runs the orchestrator and bundles no prompts fails the suite, since its runs would fail. On the recorded menu, every framework resolves a framework pack and every skill a bundled flow names. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Behaviour
An orchestrator run reads its agent prompts from the wizard, and nowhere else.
agent/. A flow with aseed: trueprompt keeps its planner. A flow without one is queued from the graph its frontmatter declares.PHW_AGENT_ORCHESTRATOR_NO_PLANNER. A failed load is also reported as an exception with the steporchestrator_bundled_prompts.resolveSkillwith the project's match keys.--context-mill-versionreads the bundled prompts like any other run. The pin selects skills only.The
wizard-agent-dagflag, the fetch ofagent-menu.jsonandagents-*.md, the runner's own menu lookup and theorchestrator_promptstag are gone. The recorded copy of context-mill's prompts underfixtures/context-mill/goes with them, andrecord:context-mill-fixturesrecords the skill menu only.Verified
pnpm typecheckandpnpm lint(0 errors).pnpm test: 291 files, 4,454 tests pass, 1 skipped.frontmatter-queue.test.tscovers the two new failures: a bundle that fails to load, and a program that bundles none.flow-graphs.test.tsandagent-prompts.test.tsstill check every bundled flow: eachdependsOnnames a task in its folder, each folder has one sink and can be ordered, and eachskillsentry is on the recorded menu.Rollout
This is the first of four stacked PRs that remove the old code paths. They merge together and ship in one release.
wizard-agent-dagandwizard-agent-dag-skillshave 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.🤖 Generated with Claude Code