Repository navigation
Conversation
Add a user-owned retro flow that turns session, branch, or pull request friction into checks and instruction changes. Teach respond to triage automated reviewer findings, settle observable forks by experiment, explain measured numbers, and write the failing test first in execute. The craft lens now applies the project's style authority before its own checklist.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Meridian should be the workflow and harness layer: the standard, model adapters, hooks, and user-owned flows. Code style should come from the project and its style skills. This change makes that split explicit. It also covers the retro and test-first workflows that previously needed outside skills.
Changes
retroflow (user-invoked, report-only). It reviews a session, branch, or PR for friction: user corrections, rework, late check failures, and review catches. It names the cause of each one and proposes the strongest durable fix, in this order: a failing check, a deleted or rewritten instruction, one new instruction line, then a skill change. Lessons that already exist as text and were still missed get a check instead of more text.house-style,golden) first, and uses its own checklist only where they are silent. Two lines no longer contradict house-style: one-call helpers are flagged only when their name adds nothing, and doc comments on exports are not slop.respondtriages automated reviewer findings. Fix when a realistic failure path exists and the fix is in proportion. Dismiss with the traced path otherwise. Ask the user before dismissing a security finding. A bot's severity label is not evidence.executewrites the failing test first when a task adds or fixes behavior and a test seam exists.KodingDev/claude-pluginsinstall. That install's session hook forces the routing skill on every request and still loads alongside a current install.retro.Not included
.agents/playbooks/withextends). Project instructions already let a repo add to a flow, so a second mechanism would duplicate layer 2.Feedback wanted
retro's fix order match how you want lessons encoded?Verification
pnpm checkpasses: format, lint, typecheck, and 39 tests. This includes the consistency test, now withretroinMANUAL_FLOWS.pnpm evaldid not run, because it makes paid model calls.Generated by Claude Code