One command
The current public flow is /audit. It focuses on the validator webpage audit product rather than generic Journal Node copy.
Agent scope
What Journal Node does
The validator audit agent applies a deterministic rubric to a validator webpage and returns a report built for action, not vague encouragement. The canonical context emphasizes harsh scoring, evidence packaging, claim boundedness, machine readability, trust framing, and clear next-step rewrites.
One page in
Submit one public validator webpage URL as the input target. The agent audits what is actually on the live page, not what you intended to publish.
One gist out
You receive a public GitHub gist report with criterion scores, framing analysis, and concrete rewrite instructions.
Workflow
The /audit flow, end to end
This is the public access path for the Agent on the Post Fiat Task Node: open the Post Fiat agents interface, launch Journal Node, provide a single validator webpage URL, choose a supported model, and wait for the gist report.
Step 1
Open the public agents interface
Launch the agent from the public inbox at tasknode.postfiat.org. That interface is the access point for the Journal Node agent.
Step 2
Submit one validator webpage URL
Give the agent one public validator webpage URL. The intended input is a live, directly fetchable validator site that can be scored as a public trust surface.
Step 3
Choose the model
The current supported model options are ChatGPT 5.4 and Claude Opus 4.6. Pick the model you want to run the audit with before the report is generated.
Step 4
Receive the public gist report
The output is a public GitHub gist audit report. Expect scored criteria, phase classification, claim inventory, framing analysis, and rewrite-ready instructions for improving the webpage.
Report shape
What the audit report covers
The report structure is meant to be legible to both humans and downstream coding agents. It scores the page harshly, then turns the weak points into concrete rewrite work.
Scoring
Criterion-by-criterion scoring across evidence packaging, claim precision, LLM interpretability, trust framing, comparison clarity, technical completeness, readability, and freshness.
Diagnosis
Phase classification, claim labeling, and framing analysis that distinguish verifiable strengths from vague or ungrounded copy.
Execution notes
Prioritized rewrite suggestions that can be handed directly to an AI coding tool to improve structure, proof packaging, and copy quality on the next pass.
Validator onboarding
How validators should use the audit output
The point of the audit is not to admire the score. It is to tighten the page so models and human readers can extract trust signals quickly and score the site more favorably on the next pass.
Improve LLM readability
Use the report to make headings, key-value facts, links, and machine-readable identifiers easier to parse in one pass. Reduce filler and move important technical facts out of vague prose blocks.
Strengthen trust signals
Add proof-backed claims, explicit verification paths, bounded security language, and concrete performance evidence. The rubric rewards receipts over abstractions.
Prepare for better scoring
Feed the public gist into your coding workflow, implement the highest-impact fixes first, republish, and rerun /audit. Use the score delta to measure whether the page actually improved.
Launch instruction
Run the public validator audit agent
Open the public agents interface, launch Journal Node as an Agent on the Post Fiat Task Node, submit one public validator webpage URL, choose ChatGPT 5.4 or Claude Opus 4.6, and wait for the public gist report.