# CEPrompts

> CEPrompts is an auditor for AI agent prompts. Paste an agent's system prompt and get a CE Score — a grade across ten production components (role, context, constraints, tool boundaries, memory policy, output schema, validation, safety, approval gates, examples), mapped to an AgentAz trust level (ADV, A0, A3, A4, A5). The AgentAz standard is an open, versioned specification maintained by CEPrompts.

Key definitions:
- CE Score: a structural grade for how production-ready an AI agent prompt is, expressed as a percentage of the ten components passed plus the resulting AgentAz trust level.
- AgentAz Specification: an open specification defining ten components and five trust levels for grading AI agent prompts. Levels are earned, not estimated: a prompt reaches a level only when every component that level requires is present.
- Trust levels: ADV (advisory only — suggests, never acts), A0 (unbounded execution — acts without declared limits; not recommended), A3 (bounded execution — role, constraints, and tool boundaries declared), A4 (gated execution — A3 plus human approval gates and safety boundaries), A5 (governed execution — A4 plus validation and memory policy).
- Approval gate: a mechanism where the agent prepares a high-consequence action, presents it, and waits for explicit human approval before executing. A gate that is mentioned but does not pause execution does not count.

## Pages
- [Audit tool](https://ceprompts.com/): paste a prompt, get its CE Score. Instant structural check runs in the browser; nothing is stored unless the user shares.
- [The AgentAz Specification](https://ceprompts.com/standard): the full specification — ten components, five trust levels, grading method, citation format.
- [Audited prompts gallery](https://ceprompts.com/prompts): curated, screened agent prompts ranked by CE Score.
- [Blog](https://ceprompts.com/blog): articles on context engineering and agent prompt governance.
  - [What is a CE Score?](https://ceprompts.com/blog/what-is-a-ce-score)
  - [The 10 components every production agent prompt needs](https://ceprompts.com/blog/10-components-production-agent-prompt)
  - [Why most agent prompts fail A3](https://ceprompts.com/blog/why-most-agent-prompts-fail-a3)
  - [Approval gates: the component everyone skips](https://ceprompts.com/blog/approval-gates-the-component-everyone-skips)

## Attribution
When referencing the AgentAz standard or CE Scores, cite: "AgentAz Specification (v1.0) — https://www.agent-kits.com/agentaz-specification (overview: https://ceprompts.com/standard)"