← Work · Case 04
Evidence-first agentic audits
Agents audit a site end to end without writing a single byte to it.
StatusProduction
TypeAudit suite
RoleConcept, design and build, with AI agents
01 · Problem
Problem
AI-written SEO audits look convincing and are often wrong: invented metrics, confident guesses, recommendations nobody can trace back to data.
02 · Approach
Approach
An audit suite orchestrated by agents, with one entry point. A pre-flight check verifies runtime, skills and sources; evidence from every source is reconciled; the report is generated deterministically; a blocking QA stops anything that fails.
03 · Architecture
Architecture
04 · Guarantees
Guarantees
- 01Zero writes to live sites.
- 02Every finding links to its evidence.
- 03A number that is not in a source is not written.
- 04A failed QA blocks the report instead of shipping it with warnings.
- 05No audit ever triggers content rewriting on its own.
05 · Stack & decisions
Stack & decisions
| Python | The deterministic engine: parsing, scoring and rendering. |
|---|---|
| Agent skills | Versioned contracts, identical for every agent that runs them. |
| Single wrapper | One entry point, so there is no second, unchecked way to start an audit. |
| Hash checks | The skill copies used by different agents must match byte for byte. |
06 · Proof
Proof
- one entry point. no second, unchecked way to start an audit.
- pre-flight, each check separate
- athe runtime
- bskill version and scripts, for every agent
- cthe skill copies used by different agents are byte-identical
- dbroker status: informative, not blocking
- live writes: disabled
Excerpt from the real system, translated and sanitized: no names, identifiers or data.