Skip to content

fix: browse binary discovery broken for agents (v0.3.5)#44

Merged
garrytan merged 4 commits intomainfrom
investigate-brokenness
Mar 14, 2026
Merged

fix: browse binary discovery broken for agents (v0.3.5)#44
garrytan merged 4 commits intomainfrom
investigate-brokenness

Conversation

@garrytan
Copy link
Owner

Summary

  • Fixed browse binary discovery — replaced find-browse indirection with explicit browse/dist/browse path. Agents were guessing bin/browse instead of running find-browse.
  • Fixed update check exit code 1 — added || true so agents don't think gstack is broken when no update is available.
  • Fixed browse/SKILL.md missing setup block$B was used but never defined in standalone /browse skill.
  • Expanded template system — added {{UPDATE_CHECK}} and {{BROWSE_SETUP}} placeholders; converted qa/ and setup-browser-cookies/ to .tmpl templates.
  • Added e2e tests — 3 Agent SDK tests for setup block (happy path, NEEDS_SETUP, non-git-repo) + LLM eval for clarity.

Pre-Landing Review

No issues found.

Eval Results

No prompt-related files changed — evals skipped.

Test plan

  • All static/validation tests pass (32 tests, 0 failures)
  • Template freshness verified (4 generated SKILL.md files match templates)
  • Manual bash test: setup block finds binary in git repo and outside git repo
  • E2E tests (gated behind SKILL_E2E=1, ~$1.50/run)
  • LLM eval (gated behind ANTHROPIC_API_KEY)

🤖 Generated with Claude Code

garrytan and others added 4 commits March 14, 2026 02:02
Agents were skipping the find-browse binary and guessing bin/browse
(wrong path). Now the setup block explicitly checks browse/dist/browse
with workspace-local priority, global fallback.

Also adds || true to update check to prevent misleading exit code 1.

Adds {{UPDATE_CHECK}} and {{BROWSE_SETUP}} template placeholders to
gen-skill-docs.ts so all skills share a single source of truth.
Replaces hardcoded update check and find-browse blocks with
{{UPDATE_CHECK}} and {{BROWSE_SETUP}} placeholders. Both skills
are now generated from templates via gen-skill-docs.
- 3 Agent SDK e2e tests: happy path, NEEDS_SETUP, non-git-repo
- LLM eval: setup block clarity + actionability >= 4
- New error pattern: 'no such file or directory.*browse'

These tests catch the exact failure mode where agents can't discover
the browse binary via SKILL.md instructions.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@garrytan garrytan merged commit 1717ed2 into main Mar 14, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant