# supabase: Apply and recover a migration

End-to-end workflows have not been measured.

Publication: 351c2b4f9f751f7b2371a4d6f8a7eed298f1ace44339a6a78015804647a77537

Task: Link the fixture to the dedicated disposable project, apply an additive schema migration, verify existing data and the new schema, replay without duplicate migration history, and recover from an invalid migration.
Limitation: Does not cover Auth, Storage, Edge Functions, or production branching. SQL state is checked by the trusted verifier, not by printed CLI success.
Measured version: Not measured
Model: Not measured

Coverage: Not measured

## Workflow outcomes

- Normal workflow: Not measured (Controls incomplete; Three valid independent trials with matching provenance required)
- Ambiguous or larger state: Not measured (Controls incomplete; Three valid independent trials with matching provenance required)
- Controlled recovery: Not measured (Controls incomplete; Three valid independent trials with matching provenance required)

## Findings

No validated workflow findings are available.

## Interface readiness

Incomplete
Mean of applicable domain scores; each domain is the fraction of applicable behavioral checks passed. All applicable checks require evidence before an overall percentage is published. No workflow outcome is included.

- Discovery and documentation: 0 of 2 measured
- Output usability: 0 of 3 measured
- Headless execution: 0 of 2 measured
- State changes, preview, and replay: 0 of 3 measured
- Errors and recovery: 0 of 3 measured
- Authentication and configuration: 0 of 4 measured

## Readiness checks

- help (discovery): unknown — Behavior has not been measured
- discovery (discovery): unknown — Behavior has not been measured
- structured-output (output): unknown — Behavior has not been measured
- clean-streams (output): unknown — Behavior has not been measured
- pagination (output): unknown — Behavior has not been measured
- help-completion (headless): unknown — Behavior has not been measured
- task-completion (headless): unknown — Behavior has not been measured
- read-safety (state): unknown — Behavior has not been measured
- preview (state): unknown — Behavior has not been measured
- replay (state): unknown — Behavior has not been measured
- invalid-command (errors): unknown — Behavior has not been measured
- invalid-flag (errors): unknown — Behavior has not been measured
- recovery (errors): unknown — Behavior has not been measured
- missing-auth (auth): unknown — Behavior has not been measured
- invalid-auth (auth): unknown — Behavior has not been measured
- valid-auth (auth): unknown — Behavior has not been measured
- scope-auth (auth): unknown — Behavior has not been measured

## Sessions


Workflow outcomes and readiness are separate. No overall CLI ranking is claimed.

## Historical cold-task

Separate from workflow journeys. Help-text probes and a one-command agent task.

AX: 61
Agent trials: 3/3
Weakest dimension: Structured I/O (30%).