CalvinBackup/gstack
1
0 Fork 0
mirror of https://github.com/garrytan/gstack.git synced 2026-10-03 18:06:54 +02:00
Code Issues Packages Projects Releases Wiki Activity

test(design): revert the three-Edit plan-mode flow

Browse Source 
A focused paid run still timed out at 300 s: the first three passes alone took
150 s of thinking. The case stays a named timeout red rather than cutting review depth.
This commit is contained in:
garrytan committed 2026-09-29 22:40:01 +00:00
1 parent a111225e78
commit 12ab3b3c54
2 files changed
+5 -6
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

test/skill-e2e-design.test.ts
+2 -2
View File
@@ -462,8 +462,8 @@ Review the plan in ./plan.md. Its design gaps are vague "clean, modern UI" and "
Use this non-interactive delivery sequence:
1. Skip the preamble bash block and any AskUserQuestion calls. Read every lazy section the workflow requires. Review all 7 design passes. Rate each scored design dimension 0-10 and explain what would make it a 10; preserve the unresolved-decisions pass and every required design decision.
2. EDIT plan.md with the missing design decisions (interaction state table, empty states, responsive behavior, etc.) and the full required review report. Save as you go in three Edits: after passes 1-3, apply their decisions to plan.md; after passes 4-7, apply theirs; then add the review report. Keep the saved review compact: use the canonical tables and decision IDs. Specify each design requirement once; refer to its section or decision ID from other pass rationales, tasks, and report cells instead of repeating that specification. Give concise score rationales and 10/10 explanations. Retain all required report fields, design decisions, diagrams, ratings, and explanations.
3. Persist that complete plan and review with those Edits (or one Write) before publishing a completed walkthrough or saying a fix is applied. Read plan.md back to verify the saved changes.
2. EDIT plan.md with the missing design decisions (interaction state table, empty states, responsive behavior, etc.) and the full required review report. Keep the saved review compact: use the canonical tables and decision IDs. Specify each design requirement once; refer to its section or decision ID from other pass rationales, tasks, and report cells instead of repeating that specification. Give concise score rationales and 10/10 explanations. Retain all required report fields, design decisions, diagrams, ratings, and explanations.
3. Persist that complete plan and review with Write before publishing a completed walkthrough or saying a fix is applied. Read plan.md back to verify the saved changes.
4. Then return a brief, concrete summary of the design changes; do not repeat the full review in the response. This changes presentation only: execute every required pass and lazy-section Read.
IMPORTANT: Do NOT try to browse any URLs or use a browse binary. This is a plan review, not a live site audit.`,