Deployment misconfiguration: In-memory mode on Vercel — new projects and imports cannot persist across servers. Set HELM_ENV_MODE=real + DATABASE_URL (+ GITHUB_TOKEN for import). See docs/VERCEL.md.

Helm

Your work across ChatGPT, Cursor, Claude & Lovable

Settings

Signal Catalog

Internal telemetry taxonomy cleanup and schema migration assistance.

Discuss

Plan what to do next in Discuss, or create a work item directly.

In Discuss you describe your goal; Helm proposes an action and you approve before anything is sent.

Open DiscussCreate work item

Project profile

Command Center

Manual control profile for AI-mediated delivery where humans retain decision control and every transition is explicit.

Governing docs

Command Center Operating Charter

docs/COMMAND_CENTER_CONSTITUTION.md

source type: operating charterprecedence: 1truth stream: rule truthrequired: true

Phase 5 Operating Rules

docs/PHASE_5_LAWS.md

source type: operating ruleprecedence: 2truth stream: rule truthrequired: true

Command Center Contract

docs/COMMAND_CENTER_CONTRACT.md

source type: operating contractprecedence: 3truth stream: audit truthrequired: true

Architecture Vocabulary

docs/ARCHITECTURE_VOCABULARY.md

source type: shared vocabularyprecedence: 4truth stream: static truthrequired: true

Phase 6.0 Design Specification

docs/PHASE_6_0_DESIGN_SPECIFICATION.md

source type: design specprecedence: 5truth stream: static truthrequired: true

Current State

docs/CURRENT_STATE.md

source type: runtime stateprecedence: 6truth stream: dynamic truthrequired: true

Lovable UI Consumption Brief

docs/LOVABLE_UI_CONSUMPTION_BRIEF.md

source type: interface guidanceprecedence: 7truth stream: interface truthrequired: false

Forbidden terms

  • best option
  • top choice
  • recommended ranking
  • highest-ranked
  • we should choose

Review checklist

  • Output confirms user remains the decision-maker.
  • No silent execution paths are described or implied.
  • AI-mediated execution includes explicit structural confirmation.
  • No recommendation/ranking language appears in reviewed Command Center outputs.
  • Document conflicts are surfaced explicitly instead of resolved by guessing.

Ambiguity handling rule

When facts conflict or are underspecified, record the conflict, mark ambiguity active, and request human clarification.

Stop conditions

  • Missing governing document with unresolved contradiction.
  • Attempted silent execution or implicit approval behavior.
  • Conflicting instructions that cannot be reconciled without human direction.
  • Insufficient structural confirmation for AI-mediated execution.