skills.rest
Noetfield-Systems
Noetfield-SystemsOFFICIAL
Explore all skills by this author →
WebsiteGitHub

sourcea-cursor-agent-knowledge-audit

Audits knowledge parity gaps between Cursor agent context and public Brain chatbot manifest.

Updated Jun 20, 2026
Search Tags:#gap analysis#manifest#ssot#cursor agent#knowledge audit#chatbot parity#eval regression
One-click install
npx skills add https://github.com/Noetfield-Systems/SourceA --skill sourcea-cursor-agent-knowledge-audit-noetfield-systems
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: sourcea-cursor-agent-knowledge-audit
Source: https://github.com/Noetfield-Systems/SourceA/tree/main/.cursor/skills/sourcea-cursor-agent-knowledge-audit
Command: npx skills add https://github.com/Noetfield-Systems/SourceA --skill sourcea-cursor-agent-knowledge-audit-noetfield-systems

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Cursor agents have full private repo context while the public Brain chatbot only sees a manifest allowlist, causing parity gaps where the agent knows answers the public chatbot cannot give. This Skill provides a repeatable audit workflow to reproduce, classify, log, and regress-test those gaps before editing agent prompts or knowledge corpora. ## Core Features & Use Cases - Gap Classification Workflow: Reproduce a question on both surfaces, then classify it as PUBLIC_SHOULD_KNOW, INTERNAL_ONLY, or STALE_PUBLIC with a defined fix path. - SSOT Read Chain: Enforces a five-file read order (manifest, positioning doc, machine SSOT, gap report, eval results) before any public-facing knowledge edits. - Eval Regression: Runs test_brain_chat_quality_v1.py against P0 buckets and appends findings to the locked gap report. - Use Case: When the founder asks "why did Cursor know X but Brain didn't?", follow the audit workflow to determine whether the gap is a pipeline issue (stale distill, missing manifest entry, unredeployed worker) rather than a model intelligence problem. ## Quick Start Ask the agent to audit why the public Brain chatbot failed to answer a specific question that Cursor answered correctly, following the gap classification workflow.

Frequently Asked Questions about sourcea-cursor-agent-knowledge-audit

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

FAQPage Schema
How do I audit why a chatbot answer differs from an AI agent answer?▼

Reproduce the same question on both surfaces, then classify the gap as PUBLIC_SHOULD_KNOW, INTERNAL_ONLY, or STALE_PUBLIC. Log the finding in the gap report and add a regression eval entry if it is a P0 issue.

How do I fix a public chatbot missing knowledge the agent has?▼

Edit the SSOT source (positioning JSON or www HTML), re-run the distill step, sync the chat knowledge bundle, redeploy the worker, then run the quality eval script and append results to the gap report.

What files should I read before editing chatbot knowledge or prompts?▼

Read the CHATBOT_KNOWLEDGE_MANIFEST.json, the positioning one-line locked doc, the machine positioning JSON, the gap report, and the last eval run results, in that order, before making public-facing edits.

Can internal repo knowledge be added to the public chatbot?▼

No. Internal-only paths, Mac ports, governance incidents, and secrets must never enter the public bundle. The chatbot should redirect politely when asked about internal-only topics like OpenRouter keys or Mac Hub ports.

Why does the chatbot still fail after updating its knowledge source?▼

Check whether the distilled output is newer than the source, whether the sync step ran, and whether the worker was redeployed after the bundle change. If any step was skipped, the gap is a pipeline issue, not a model problem.

Related Skills

Explore More Software Engineering Skills→
Kingly-AgencyKingly-Agency

rag-implementation

Build RAG pipelines that ground LLM responses with retrieved external knowledge.

Official
Advanced
ThalesGroupThalesGroup
19

agilab-docs

Synchronize AGILAB canonical docs sources with public mirrors and validate changes.

Official
Intermediate
rivaldiekaptrrrrivaldiekaptrrr

eval-harness

Define and run eval suites to validate AI task performance.

Community
Advanced
JennyfirrrJennyfirrr

workspace-lint

Audit workspace structural contracts and report drift as pass/fail.

Community
Advanced