Files
crank/docs/capability-baseline/manual-checklist.md
T

4.1 KiB

Community UI capability baseline checklist

baseline_version: 2026.08.24.1

This is a bounded brownfield baseline, not the release-candidate regression from Epic 8. For every charter record happy, loading, empty, error, recovery, stale-response, RU/EN, and safe-output observations. Use not_run for an applicable unexecuted state, gap for a missing required contract, and n/a only with a by-design reason.

UI-01 Authentication and workspace

  • flow_id: ui-auth-workspace
  • states: happy, loading, error, recovery, stale, ru-en, safe-output
  • verdict: pass
  • reason: Wrong-password error and successful recovery were observed; settings and RU/EN navigation remained usable.

UI-02 Operations catalog and lifecycle

  • flow_id: ui-operation-lifecycle
  • states: happy, loading, empty, error, recovery, stale, ru-en, safe-output
  • verdict: fail
  • reason: Happy, loading, empty, error and recovery rendered, but an older response overwrote newer state and a credential-shaped error canary reached the DOM; see DEF-UI-001 and DEF-UI-003.

UI-03 Operation test execution

  • flow_id: ui-operation-test
  • states: happy, loading, error, recovery, stale, ru-en, safe-output
  • verdict: pass
  • reason: Wizard happy path, validation feedback and recovery controls were observed; loading and empty are not applicable to this local step flow.

UI-04 YAML and OpenAPI import/export

  • flow_id: ui-operation-import-export
  • states: happy, loading, empty, error, recovery, stale, ru-en, safe-output
  • verdict: fail
  • reason: Import entry and recovery are present, but EN mode still exposes Russian-only OpenAPI text; see DEF-UI-002.

UI-05 Agent lifecycle and bindings

  • flow_id: ui-agent-management
  • states: happy, loading, empty, error, recovery, stale, ru-en, safe-output
  • verdict: fail
  • reason: Happy, error and recovery rendered, but stale-response protection is absent and a credential-shaped error canary reached the DOM; see DEF-UI-001 and DEF-UI-003.

UI-06 MCP and approval keys

  • flow_id: ui-mcp-approval-keys
  • states: happy, loading, empty, error, recovery, stale, ru-en, safe-output
  • verdict: pass
  • reason: Agent selection, key creation entry and single raw-key disclosure guidance were observed; the seeded stack made the initial empty state inapplicable.

UI-07 Auth Profiles, Upstreams, and Secrets

  • flow_id: ui-auth-profiles-upstreams-secrets
  • states: happy, loading, empty, error, recovery, stale, ru-en, safe-output
  • verdict: fail
  • reason: Happy, real empty, error and recovery rendered, but a credential-shaped error canary reached the DOM; see DEF-UI-003.

UI-08 Logs and approvals

  • flow_id: ui-logs-approvals
  • states: happy, loading, empty, error, recovery, stale, ru-en, safe-output
  • verdict: fail
  • reason: Happy, error and recovery rendered, but a credential-shaped error canary reached the DOM; see DEF-UI-003.

UI-09 Usage

  • flow_id: ui-usage
  • states: happy, loading, empty, error, recovery, stale, ru-en, safe-output
  • verdict: fail
  • reason: Happy, error and recovery rendered, but a credential-shaped error canary reached the DOM; see DEF-UI-003.

UI-10 Cross-screen stale response

  • flow_id: ui-auth-workspace
  • states: stale, recovery
  • verdict: fail
  • reason: A delayed older Operations response overwrote a newer empty response; see DEF-UI-001.

UI-11 Localization consistency

  • flow_id: ui-operation-lifecycle
  • states: ru-en
  • verdict: fail
  • reason: Switching to EN left RU fragments on the Operations screen and OpenAPI import has hardcoded RU text; see DEF-UI-002.

UI-12 Unsafe output canaries

  • flow_id: ui-logs-approvals
  • states: safe-output, error
  • verdict: fail
  • reason: Credential-shaped API error canaries were visible in multiple browser error states; see DEF-UI-003.

UI-13 Getting Started to authoritative first Tool call

  • flow_id: ui-onboarding-first-call
  • states: happy, loading, empty, error, recovery, stale, ru-en, safe-output
  • verdict: not_run
  • reason: The five-participant fresh-install usability protocol is documented, but no participant run or clean-revision evidence has been collected. Do not infer a ≤15-minute median, completion rate, or secret-safety observation from automated source changes.