LLM Squared Knowledge Review Analyst 1.0.0

What it does
Checks explicitly supplied questions, answers, source versions, citation records
and completed human-review declarations. It produces a complete review packet
with unresolved items and retest requirements. It does not generate policy
answers, verify a person, approve an institution or connect to any account.

Start with the browser: https://llmsquared.com/tools/knowledge-acceptance/
Use only permitted information. Exports retain every supplied input.

Local command
A currently supported Node.js runtime with Web Crypto is required.
No package installation or outside service is used by this command.

node cli.mjs --help
node cli.mjs review fictional-input.json --output my-review.json
node cli.mjs review fictional-input.json --format html --output my-review.html

The included fictional-review.json and fictional-review.html are reference
outputs. The commands above create new files beside them on a fresh extraction.

Existing output files are preserved unless --overwrite is explicitly supplied.
Input-only files saved from the browser are accepted. Imported complete reports
are not treated as input: save the input draft separately in the browser.

Assistant use
The knowledge-review-analyst/ folder contains a focused SKILL.md and UI metadata.
Review it under your approved assistant setup process. A skill is an instruction
resource; it does not install a server or establish actual model invocation.
If an assistant reads private input, its host may transmit it to a model provider.
Use an institution-approved provider and handling route.

Context fingerprints only bind a declared review to supplied text and metadata.
They do not establish authorship, source authenticity, truth or authorization.
The institution decides acceptance after real qualified human review.

This separate package does not add a tool to Evidence Desk 0.6.0 or establish
Claude, ChatGPT, Gemini or Grok directory acceptance. No remote MCP is exposed.
