Google Drive Organization
Organize Google Drive folders without breaking permissions, links, shortcuts, or retention rules. Includes duplicate decisions, staged moves, and verification.
givaro91
Updated Jul 28, 2026
Design and release Lark Base workspaces without losing control of schema, relationships, permissions, migrations, or workflow side effects. Includes staged verification and ambiguous-write recovery.
A Lark Base can look tidy while its underlying process is brittle. Duplicate customer records split history. A status workflow triggers itself. A replay creates two tasks. A hidden field remains visible through permissions. A migration reports the right count but links records to the wrong owner.
Lark Base Automation starts with the business process, stable Base and resource IDs, source-of-record owners, data sensitivity, coverage, and separate authority for design, schema changes, imports, permissions, workflows, notifications, and deletion. It keeps observed configuration, proposed design, approved change scope, live effects, and verification evidence separate.
The workflow produces an entity map, field dictionary, relationship map, lifecycle model, role-specific surfaces, and one contract per automation. Each automation names its trigger, preconditions, duplicate-prevention key, effects, required non-effects, actor, recipients, limits, failure owner, tests, and disable or rollback path. It explicitly checks for self-triggering workflows, cycles, duplicate tasks, unintended recipients, and private-data leaks.
Migrations use immutable source exports, row-level source-to-target mappings, explicit exception classes, a reversible pilot, bounded batches, and independent read-back. A timed-out or incomplete operation is reconciled target by target as an exact match, no effect, partial effect, conflict, multiple matches, or unavailable evidence. Any repair gets a new operation ID and is bound to the current schema, exact relationship target, record revision, workflow state, and duplicate key before another write is considered.
The package includes a reusable Lark Base contract for data models, surfaces, automation controls, migrations, change sets, operation ledgers, verification, and handoff. It can guide or review live work, but it cannot authenticate Lark, grant permission, approve a business rule, choose a source of record, mutate a Base, or verify production behavior without the required access, evidence, and authority.
Version 1.0.0 was evaluated on 29 July 2026 through Codex CLI with gpt-5.6-sol in 12 fresh read-only contexts. Across three blind comparison cases, the candidate scored 100, compared with 85.42 for the same agent without the skill and 85 for the exact free upstream. It passed every case and led the aggregate comparison by 14.58 and 15 points. The cases covered a customer-service Base design, a partially evidenced leave-management release, and recovery from an ambiguous cross-Base sync. One migration case tied the upstream at 100. This evaluation measures the supplied cases; it does not prove that a real Lark workspace was changed.
The reviewed package SHA-256 is 22b5f4c5ae4cbee337f8e9244fb0582a112fcd8325eae6666f190ba407b590d8. Maintenance and install support are handled by the platform-operated publisher for this listing.
Input
We handle Appointments, Authentications, Buybacks, and Sales. Design Customers, Requests, and Products from this export. Only a confirmed Sale may create one Shopify order. Frontline staff must not see margins. Design only.
Output
A source-bound entity and field model, lifecycle states, relationship rules, role-specific surfaces, one bounded automation contract per request type, migration exception classes, pilot and verification gates, required non-effects, and a design-only handoff.
Input
The 20-row pilot returned success, but we have no row mapping, relationship checks, permission read-back, or fixture results. The workflow runs on every Status change and writes Status again. Can we release the remaining 988 rows?
Provide the business process, Base URL or stable IDs, source systems, existing tables and fields, sample records, desired views or forms, workflow rules, roles, privacy and retention constraints, migration evidence, and separate authority for inspection and each type of change. Ask for a Base design, automation contract, migration review, staged release plan, timed-out operation reconciliation, verification matrix, or operational handoff.
Business process, owner, users, included and excluded work, deadline, expected outputs, service expectations, and success criteria.
Base URL or stable app token, table and resource IDs, paginated schema and record inventory, field types, relationships, views, forms, workflows, roles, permissions, and integrations when available.
Source systems, stable source keys, source-of-record owners, sample or export revision, sensitivity, retention, time zone, volume, validation rules, and known exceptions.
Separate permission to inspect, design, create, import, update, delete, change access, enable workflows, notify people, publish forms, or call external systems.
SKILL.md; Lark Base design, automation, migration, recovery, verification, and handoff contract; Codex interface metadata
No reviews yet.
Output
A blocked release verdict, distinct duplicate and invalid-row queues, row-level pilot reconciliation, a corrected idempotent workflow contract, representative fixtures, new approval gates, rollback or repair rules, and a partially applied state.
Input
One source has exactly one correct destination and task. One destination is missing its Owner and task. A third source has two partial destinations and two tasks. Do not mutate anything; tell us what is safe.
Output
Exact-match, partial-effect, and multiple-match verdicts; workflow containment; no replay of completed work; current-schema and exact-target gates for a new repair operation; an owner decision for duplicates; non-effect checks; and a blocked handoff.
Creator
Ssavrio86