Show WebVM target artifacts in VS Code Explorer

[konard]

Fixes #25.

Summary

• Keep /workspace/target visible in the VS Code Explorer by syncing target artifact metadata instead of pruning the directory.
• Store target artifacts as metadata-only IndexedDB entries so Explorer can show the tree without copying generated binaries/intermediates into the browser workspace.
• Align the generated sync hook and baked disk bash profile, add regression coverage, and replace the issue Improve transparency between VS Code UI and WebVM, so it works naturally and as user expects. #25 case study with evidence, screenshots, logs, and research notes.

Root Cause

The guest sync hook introduced for issue #21 used this traversal:

find /workspace -path /workspace/target -prune -o -path /workspace/.git -prune -o -print

That kept prompt-time sync fast, but it also made the browser workspace incomplete. Cargo created target/ in the WebVM, but VS Code never received tree entries for it.

Fix

• web/glue/workspace-sync.js now traverses /workspace/target and emits target files as skipped metadata records with sizes.
• web/glue/workspace-fs.js now supports metadata-only files, shows them through stat()/readDirectory(), rejects content reads with Unavailable, and excludes them from WebVM priming snapshots.
• web/disk/Dockerfile.disk now bakes the same sync protocol into /root/.bash_profile.
• web/tests/workspace-sync.test.mjs reproduces the old prune behavior and verifies target metadata visibility/refresh.
• web/tests/e2e/local-pages-e2e.test.mjs asserts that the warm-disk workspace tree includes target when browser e2e runs.

Reproduction Evidence

• docs/case-studies/issue-25/verification/focused-before.log shows the focused regression test failing before the fix because /workspace/target was still pruned.
• docs/case-studies/issue-25/screenshots/local-target-visible.png is the reporter's local VS Code reference.
• docs/case-studies/issue-25/screenshots/rust-web-box-target-hidden.png is the reporter's rust-web-box failure screenshot.

Verification

• node --test web/tests/workspace-sync.test.mjs passed: 8 tests.
• node --test web/tests passed: 179 passed, 3 skipped because local Playwright/browser-commander or live URL was not configured.
• Generated sync hook passed bash -n.
• git diff --check passed.
• cargo fmt --check passed.
• cargo clippy --all-targets --all-features passed.
• cargo test --all-features --verbose passed.
• cargo test --doc --verbose passed.
• rust-script scripts/check-file-size.rs passed.
• cargo build --release --verbose passed.
• cargo package --list --allow-dirty passed.
• rust-script scripts/check-changelog-fragment.rs passed with PR-like env vars.
• rust-script scripts/check-version-modification.rs passed with PR-like env vars.

Notes

Target file contents are intentionally not mirrored into IndexedDB. Opening a metadata-only target artifact from Explorer returns Unavailable rather than fake file contents, while cargo run continues to use the real files in the WebVM filesystem.

[konard]

Working session summary

Implemented and pushed the fix to PR #26: #26

What changed:

• /workspace/target is no longer pruned from guest-to-browser sync.
• Target artifacts now appear in VS Code Explorer as metadata-only files with sizes, avoiding large artifact body sync.
• Metadata-only entries are excluded from WebVM priming snapshots, so fake target files are not written back into the VM.
• Updated generated sync hook, baked disk hook, regression tests, e2e assertion, changelog, and issue Improve transparency between VS Code UI and WebVM, so it works naturally and as user expects. #25 case study evidence.

Verification:

• Local focused regression: node --test web/tests/workspace-sync.test.mjs passed, 8 tests.
• Local full web tests: node --test web/tests passed, 179 passed and 3 skipped due missing local browser/live-url setup.
• Rust checks passed: fmt, clippy, tests, doc tests, release build, package list, file-size, changelog/version policy checks.
• Fresh CI on SHA 8d7a4f7 passed:
  • CI/CD Pipeline 25592287951, created 2026-05-09T04:58:06Z
  • Build disk image 25592287800, created 2026-05-09T04:58:05Z
  • Deploy GitHub Pages 25592287799, created 2026-05-09T04:58:05Z
• PR is ready for review and mergeable cleanly.

Research sources recorded in the case study:
Cargo build cache: https://doc.rust-lang.org/cargo/reference/build-cache.html
Cargo incremental profiles: https://doc.rust-lang.org/cargo/reference/profiles.html#incremental
VS Code Explorer behavior: https://code.visualstudio.com/docs/getstarted/userinterface#_explorer-view
CheerpX filesystem docs: https://cheerpx.io/docs/guides/File-System-support

---

This summary was automatically extracted from the AI working session output.

[konard]

🤖 Solution Draft Log

This log file contains the complete execution trace of the AI solution draft process.

💰 Cost estimation:

• Model: GPT-5.5
• Provider: OpenAI
• Public pricing estimate: $17.150118

📊 Context and tokens usage:

• 532.1K / 1.1M (51%) input tokens, 46.1K / 128K (36%) output tokens

Total: (532.1K + 9.8M cached) input tokens, 46.1K output tokens, $17.150118 cost

🤖 Models used:

• Tool: OpenAI Codex
• Requested: gpt-5.5
• Model: GPT-5.5 (gpt-5.5)

📎 Log file uploaded as Repository (56381KB)

• View complete solution draft log

---

Now working session is ended, feel free to review and add any feedback on the solution draft.

[konard]

✅ Ready to merge

This pull request is now ready to be merged:

• All CI checks have passed
• No merge conflicts
• No pending changes

---

Monitored by hive-mind with --auto-restart-until-mergeable flag