This commit addresses the issue where DeepSeek's raw XML markup (like <tool_calls> and <think>) was leaking into chat history, causing hallucinations in subsequent turns. It also patches a vulnerability in the git commit tool where arbitrary shell injection was possible.
Additionally, it includes UX copy and color contrast adjustments for the marketing homepage breadcrumbs.
Phase 6 captures the Bolt parity work that came out of today's
review of Bolt screenshots — two-pane shell, preview-as-default,
plan-as-checklist, settings popover, project-level secrets.
Adds a "Model assignment convention" section so we can
explicitly route mechanical work to a cheaper model and reserve
Opus for judgment-heavy tasks. Each Phase 6 row tagged opus /
cheap / opus-spec→cheap. Net: 9 hrs Opus, 8 hrs cheap.
Also brings forward two items shipped today that weren't in
the plan yet:
- 5.7 dev container <-> Gitea wiring (auto-clone +
auto-commit + GITEA_USERNAME fallback fix)
- 3.8a/b/c "stop at something tangible" rule + reverted
composer chip row + queued server-side enforcement
Sequencing diagram + cadence note updated to include P6.
Co-authored-by: Cursor <cursoragent@cursor.com>
Marks done in BETA_LAUNCH_PLAN.md:
2.4 — Coolify deploy-failed → Slack
2.9 — Sentry-as-product loop (all 4 stages)
3.4 — URL chips: +N popover
3.5 — Status pill: deep-link to Coolify
4.6 — Per-workspace soft caps (3 projects + 3 dev containers)
5.1 — vibn-dev:latest image healthy on Coolify host
Adds detailed smoke-test runbook (10 steps) for task 4.1, the
last open item before invite-1. Each step has a "Verify" line
naming exactly which subsystem it exercises (Sentry, quotas,
URL chips, status pill, Slack) so a single run covers the
entire Phase 2 + 3 + 4 + 5 surface that shipped today.
Bumps vibn-frontend submodule with the implementation work.
Co-authored-by: Cursor <cursoragent@cursor.com>
The Sentry wiring shipped today catches errors in the platform
itself. The bigger product opportunity is wiring Sentry into every
project Vibn ships, then feeding those errors back into the user's
AI chat — turning the AI from "codes for you" into "owns the
product."
Proposal lays out 4 staged shippable chunks (~8 hr total):
1. Auto-provision per-project Sentry on first deploy
2. Bake Sentry into every scaffold template
3. Expose project_recent_errors / _detail / _resolve as MCP tools
4. Auto-surface unresolved errors at chat-turn start
Folded into BETA_LAUNCH_PLAN.md as Phase 2.9.
Co-authored-by: Cursor <cursoragent@cursor.com>
- 2.6 (Dockerfile HEALTHCHECK 127.0.0.1) already in place; marked done.
- 2.7 audit: vibn-dev and vibn-agent-runner have no HEALTHCHECK so
cannot hit the localhost/IPv6 trap. Marked done.
- 2.8 (NEW): tool-error recovery middleware shipped — bumps the
vibn-frontend submodule to c105b42.
Co-authored-by: Cursor <cursoragent@cursor.com>
