Repository navigation
Release: v1.7.1 - #25
Merged
Merged
Conversation
… default Releases had stopped being tagged (v1.4.1 in July, then six PRs with no tag until the backfill). Wire the JarvusInnovations/infra-components release-prepare / release-validate / release-publish workflows so pushing develop opens a Release PR and merging it tags and publishes. main stays the default branch because `npx skills add` installs from the default branch and a fresh install should get the latest release, not unreleased work. Two small workflows cover the one cost of that choice: - main-guard: a required check on PRs into main that fails unless the head is develop (the Release PR) or hotfix/*, and always fails for forks. - main-retarget: on open, moves any other PR targeting main onto develop via the API and leaves a comment. pull_request_target with no checkout. check.yml now also runs on pushes to develop. README gains a short Contributing and releases section. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
ci(release): adopt the develop→main Release-PR flow with main kept as default
Author
Changelog- ci(release): adopt the develop→main Release-PR flow with main kept as default [#24] @themightychris
- docs(specops): builders run on an explicitly passed builder-tier model, never the orchestrator's [#32] @themightychris |
…l, never the orchestrator's The protocol said "use Sonnet for the implementation subagents" but did not say how, and the harness default is to inherit the orchestrator's model when a dispatch omits `model`. An orchestrator on the strongest tier then runs every builder on that tier: the most expensive way to implement, and one usage-limit event on that tier stops every builder at once. State the rule as a dispatch requirement: pass `model` on every Agent call (Sonnet by default, Opus for correctness-critical plans), set CLAUDE_CODE_SUBAGENT_MODEL as the backstop, and reserve the orchestrator's own model for the human's explicit request. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015AK5HiZZnymqkm7nVLfnpP
… the same builder The adversarial pre-review pass runs first on every branch; the orchestrator then reads the diff with that report in hand and owns the merge. Each finding is routed: a substantive fix goes back to the implementation subagent that built the branch, resumed so it keeps its context, never a fresh builder; a quick, unambiguous fix the orchestrator makes directly. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015AK5HiZZnymqkm7nVLfnpP
…icit docs(specops): builders run on an explicitly passed builder-tier model, never the orchestrator's
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adopts the develop→main Release-PR flow in this repo, with
mainkept as the default branch sonpx skills addkeeps installing the latest release, and tightens the drain-the-DAG protocol: implementation subagents run on an explicitly passed builder-tier model, never the orchestrator's, and the review gate is two passes (adversarial subagent first, then the orchestrator's own read) with each finding routed back to the same builder or fixed directly.Technical