CRUD Completeness Audit for Every Entity
The Idea
For every entity in your system, verify the agent has the full four — create, read, update, delete:
- Create. Can the agent make new instances?
- Read. Can the agent see what exists?
- Update. Can the agent modify instances?
- Delete. Can the agent remove instances?
The audit is mechanical: list every entity, verify all four operations are available. Do it on a checklist; don’t trust memory.
The Most Common Failure
You build create_note and read_notes because those are the first features anyone needs. You forget update_note and delete_note because individually they feel less urgent. Then a user asks the agent to “fix that typo in my meeting notes,” and the agent has no way to help. The capability gap doesn’t surface until the moment a user runs into it.
Why It Matters
CRUD audits are the cheapest parity insurance. Every gap they catch is a request the agent would otherwise have failed silently — and silent agent failures erode trust faster than visible ones do.
Related
- Parity - Agents Need Tools for Everything the UI Can Do — the audit is the operational version of parity
- Entity-Scoped Directories for Agent-Native Apps — the entities in the directory shape are the same entities to audit
- Granularity - Tools Are Atomic Primitives, Features Are Outcomes — CRUD primitives are the atomic unit of entity access
- Dynamic Capability Discovery Over Static Tool Mapping — discovery and CRUD interact: with discovery, CRUD coverage falls out of the API
- Agent-Native Anti-Patterns - Router, Workflow Tools, Defensive Design — missing CRUD is the most common form of “orphan UI actions”