Merged
Conversation
added 10 commits
February 13, 2026 21:12
Adds a new verification phase between diagnose and fix that researches whether reported behavior is an actual bug or intended design. This prevents wasting effort attempting fixes for non-bugs.
The withastro/compiler repo may be cloned as a sibling directory. Instructs the diagnose skill to check it when stack traces point to compiler behavior, and the fix skill to document proposed compiler changes in report.md.
Adds a new step to verify browser/runtime compatibility before implementing a fix. Also adds compiler repo awareness and renumbers steps accordingly.
Adds verifyResultSchema and runs verify after diagnose. Skips the fix step when verification determines behavior is intended. Also clones the compiler repo in CI so diagnose/fix skills can reference it.
Adds explicit 'read report.md' to the critical instruction at the top of every skill file, so agents always load prior context before appending their own findings.
The bullet list was redundant — the agent reads report.md and extracts whatever context it needs. Replace with a single instruction.
Extracts shouldRetriage, selectTriageLabels, fetchIssue, and runTriagePipeline from the monolithic triage function. Inlines schemas next to their call sites, adds early returns for non-reproducible and intended-behavior cases, and validates the issue response with valibot.
|
added 10 commits
February 13, 2026 22:32
This reverts commit ef0ea98.
Princesseuh
approved these changes
Feb 14, 2026
added 6 commits
February 14, 2026 15:53
* fix(markdoc): sync custom attributes between tags and nodes with shared names
In Markdoc, `table` exists as both a tag (`{% table %}`) and a node
(the inner table structure). When users configure custom attributes on
`nodes.table` or `tags.table`, the AST propagates those attributes to
both the tag and node, but validation only checks the schema for each
type independently. This caused "Invalid attribute" errors when
attributes were declared on only one side.
Add `syncTagNodeAttributes()` to automatically merge attribute
declarations between tags and nodes that share the same name after
config setup, so users can define attributes on either side.
Fixes #14220
* chore: clarify why explicit types are needed on builtinTags/builtinNodes
…m sandbox The sandbox subagents had no access to issue data because the orchestrator fetched it but never passed it through. This caused the reproduce skill to attempt gh CLI calls which fail without a token in the sandbox. - Extract IssueDetails valibot schema and type in issue-triage.ts - Fetch additional fields (author, labels, state, authorAssociation, etc.) - Pass issueDetails as args to reproduce, diagnose, verify, and fix skills - Add issueDetails prerequisite to diagnose, verify, and fix skill docs - Replace gh CLI commands in sandbox-run skills with curl/git alternatives - Use author_association field for maintainer detection instead of gh api - Remove gh CLI from sandbox Dockerfile (not usable without token)
…to comment skill The comment skill now receives available priority labels and selects one as part of rendering the comment. This makes the priority judgment visible in the posted comment (answering 'how bad is it?') and lets the downstream label selector simply extract it rather than deciding independently. - Extract fetchRepoLabels helper with valibot validation - Pass priorityLabels to comment skill, packageLabels to label selector - Simplify selectTriageLabels to extract priority from comment + pick packages
- Add assert helper for runtime invariant checks - Validate flue.args with valibot instead of unsafe type cast - Add v.nonEmpty() to label result schema, remove dead null guard - Assert fetched label arrays are non-empty before proceeding
This was referenced Feb 15, 2026
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.
Lots of small improvements as we finally start to get some broken / incomplete triage runs back from the auto-triage bot.
Summary
verifystep to the triage pipeline that checks whether reported behavior is intentional before attempting a fix. Fixes issues where the bot just trusted the submitting user's expected behavior as truth vs. potentially confused/incorrect on expected behavior.Testing
No good way to test CI locally, so will need to test a bit post-merge.