# RepoChicken: AI Repository Context Audit

> Preview of a scoped audit service for engineering teams already using coding agents.

## What is a repository context audit?

A repository context audit compares agent instructions and supporting knowledge with the repository they describe. It identifies supported defects such as obsolete commands, conflicting rules, missing discovery paths, and unclear instruction scope. It separates confirmed findings from hypotheses needing a live agent test.

The proposed engagement covers one repository, one agreed agent environment, and up to three representative workflows.

## Who is it for?

Engineering leaders and repository owners whose teams use GitHub Copilot, Claude Code, Codex, or similar coding agents and repeatedly correct repository-specific mistakes. Support must be assessed in the customer's actual configuration; the proposal does not guarantee equivalent behavior across tools.

## What is delivered?

- A context inventory: entry points, scope, knowledge routes, and open validation questions.
- An evidence-backed findings report with file references, consequences, priority, and confidence.
- A prioritized remediation plan with acceptance checks.
- A technical review session to discuss ownership and next steps.

Implementation, controlled before/after agent trials, security auditing, and code-quality auditing are not included in the proposed scope.

## Prototype terms

Illustrative price: USD 2,500 once, not a subscription.
Illustrative turnaround: five business days after access.
Proposed involvement: a 30-minute intake and a 45-minute review.
These assumptions have not been validated through customer engagements. No booking or payment is available.

## Access and privacy

A real engagement would agree read-only repository access or an export, permitted tools and model providers, command execution permissions, retention, and deletion in writing. This prototype accepts no repository uploads. Missing or redacted material limits conclusions. Any future case study requires separate explicit permission.

## Next step

Send inquiries to hello@repochicken.com. The [audit page](/#request) contains the prefilled email link; no purchase or booking is completed by visiting it.

The intake asks for company name, engineering team size, current coding-agent environment, a repository description without private code or credentials, and recurring agent mistakes. A real engagement would agree scope before work begins.

## Why not just generate AGENTS.md?

Generated instructions still need to be checked against the code, configuration, applicable scope, and other documentation. The audit's value is evidence and actionable corrections. Adopting OKF or buying a new platform is not required.

## Does less context prove better performance?

No. Smaller files do not establish lower billed tokens, faster task completion, or better outcomes. Those conclusions require controlled and repeated task runs. No such improvement has been measured in this prototype.

## Evidence and related reading

- [Repository context guide](/repository-context.md): Key concepts and evidence standards.
- [Demonstration audit](/sample-audit.md): Our own repository at commit 3b2678d; two confirmed issues and one validation gap.
- [Readable report](/sample-audit.html): The complete report with a download.

This is a first-party demonstration, not customer results.