audit-grounding

Compare diffs against CONTEXT.md, docs/adr, and knowledge base for semantic divergences.

4|Updated Mar 21, 2026

Search Tags:#audit #grounding #source-of-truth #diff-audit #invented-convention #unsupported-domain-fact #semantic-quality

Ready to run

npx skills add https://github.com/leocosta/octopus --skill audit-grounding

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: audit-grounding
Source: https://github.com/leocosta/octopus/tree/main/skills/audit-grounding
Command: npx skills add https://github.com/leocosta/octopus --skill audit-grounding

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Confronts diffs with the team's living source of truth to surface semantic mismatches, such as invented conventions and unsupported-domain claims, without blocking work. It reads core references in CONTEXT.md, docs/adr, and the knowledge base, then reports findings for human review.

Core Features & Use Cases

Load the source of truth in order (CONTEXT.md, docs/adr/*, knowledge base) to evaluate diffs.
Detect invented conventions and unsupported-domain facts, flagging them for review while never blocking tasks.
Provide signal-level findings that guide reviewers during reviews, merges, or QA tasks.

Quick Start

Run the grounding audit against the current diff to surface invented conventions and unsupported-domain facts.

Frequently Asked Questions about audit-grounding

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I audit code diffs against my project's source of truth?▼

You can audit code diffs by comparing working tree, feature branch, or pull request changes against CONTEXT.md, docs/adr, and the knowledge base to surface semantic divergences for human review.

What is a diff audit for semantic grounding?▼

A diff audit for semantic grounding identifies semantic divergences between code changes and the living source of truth. It detects invented conventions and unsupported domain facts without blocking work, emitting signal-level reports for review.

How do I detect invented conventions in a pull request?▼

Detect invented conventions in a pull request by running a grounding audit that compares changed files against CONTEXT.md, docs/adr, and the knowledge base, emitting warning signals for unsupported claims.

Can I run a semantic audit on a feature branch without blocking the merge?▼

Yes, you can run a semantic audit on a feature branch without blocking merges. The audit reports findings as info or warn signals for human review without enforcing hard blocks on the workflow.

What do I need to set up before checking diffs for unsupported domain facts?▼

Before checking diffs for unsupported domain facts, ensure your project contains a living source of truth. The audit loads references in order from CONTEXT.md, docs/adr, and the knowledge base to evaluate changes.

When should I avoid using an automated diff audit?▼

You should avoid relying on an automated diff audit as a gate, because it is designed to surface findings for human review without blocking work. It provides signal-level reports rather than enforcing merge constraints.