Debug: Debug Dangerfile PR data content#34398

Closed

Sidnioulz wants to merge 49 commits into

project/sb-agentic-setupfrom

Commits on Mar 30, 2026

Add LLM eval system for Storybook agentic setup (M0)
kasperpeulen
committed
Migrate eval agents from CLI to SDK
kasperpeulen
committed
Incorporate improvements from PR review + pre-prepared repos
kasperpeulen
committed
Add Google Sheets upload, run IDs, and environment capture
kasperpeulen
committed
Add composable prompt variants and vitest-based self-heal
kasperpeulen
committed
Simplify eval codebase (-308 lines)
kasperpeulen
committed
Remove cleanEnv from grading — only needed for installDeps
kasperpeulen
committed
Remove cleanEnv entirely — .npmrc is only in the monorepo, not in trial dirs
kasperpeulen
committed
Switch from jiti to native Node TS support, add .ts extensions to all imports
kasperpeulen
committed
Update models: Sonnet 4.6, Opus 4.6, Haiku 4.5, GPT 5.4 Medium/High
kasperpeulen
committed
Decouple agent × model × effort as three independent axes
kasperpeulen
committed
Simplify prompt to single name, add per-agent default model
kasperpeulen
committed
Split into eval.ts (single run) and eval-parallel.ts (8 runs)
kasperpeulen
committed
Add prefixed logging for parallel runs
kasperpeulen
committed
Spawn separate node processes in eval-parallel for multi-core CPU usage
kasperpeulen
committed
Live-stream prefixed logs from child processes, improve Codex agent logging
kasperpeulen
committed
Fix Codex agent logging to match actual SDK event/item types
kasperpeulen
committed
Decouple agent and model — choose agent then model independently
kasperpeulen
committed
Clean up names: claude-code→claude, claude-sonnet-4-6→sonnet-4.6
kasperpeulen
committed
Infer agent from model — node eval.ts -m gpt-5.4 auto-selects codex
kasperpeulen
committed
Fix parallel race condition: add prompt + random suffix to trial IDs
kasperpeulen
committed
Use crypto.randomUUID for unique trial IDs
kasperpeulen
committed
Fix ghost stories to match core implementation: pass paths as args + env, 1s timeout, no --project
kasperpeulen
committed
Add tests, import ghost stories utilities from core, switch to parseArgs
kasperpeulen
committed
Simplify eval harness: merge config, options objects, remove duplication
kasperpeulen
committed
Fix: stop importing parse-vitest-report from core (extensionless imports)
kasperpeulen
committed
Add .ts extensions to core imports used by eval harness
kasperpeulen
committed
Fix ghost-stories comment to reflect inline vitest parsing approach
kasperpeulen
committed
Use parseVitestResults from core for ghost stories grading
kasperpeulen
committed
Refactor eval harness: injectable logger, Node IPC, parallel grading, tsconfig fixes
kasperpeulen
committed
Refactor ghost stories: rename runStoryTests→runGhostStories, export from core-server, inline into grade.ts
kasperpeulen
committed
Fix native Node TS execution: use direct file imports with .ts extensions
kasperpeulen
committed
Simplify eval harness: Promise.all, async fs, remove Google Sheets, drop exec wrapper
kasperpeulen
committed
Refactor eval system: Zod schemas, unified CLI, shared utilities
kasperpeulen
committed
Update AGENTS.md and tsconfig comments for native Node TS execution
kasperpeulen
committed
WIP: checkpoint current eval harness changes
kasperpeulen
committed
Fix eval ghost-stories globbing lint
kasperpeulen
committed
Refine eval grading review fixes
kasperpeulen
committed
Rewrite review-pr skill: scrollable single-page instead of Reveal.js slides
kasperpeulen
committed
Update review-pr skill: two-layer format, TS-highlighted diffs, readability review
kasperpeulen
committed
Update review-pr skill: show full interface bodies in walkthrough
kasperpeulen
committed
Refactor: use composition for AgentRunConfig instead of extends
kasperpeulen
committed
Rename eval data structures for clarity
kasperpeulen
committed
Make Project.branch required and remove non-null assertion
kasperpeulen
committed
Refine eval API: discriminated AgentVariant union, rename runTask→runTrial, throw on ghost story errors
kasperpeulen
committed
Fix CI: format eval files, fix effort type narrowing in Claude agent
kasperpeulen
committed
DEBUG DANGERFILE
Sidnioulz
committed
debug
Sidnioulz
committed
foo
Sidnioulz
committed