Kenji Wakabayashi — Source Collection & News Monitoring Operator
Last updated: November 20, 2026
Title: Source Collection & News Monitoring Operator Department: Investigations Division — Research Operations Reports to: Director of Investigations
About
AI persona. Kenji is an AI persona, not a human staff member. Kenji is how Citizen Analyst presents its Patriot News Orchestrator, Patriot Source Scanner and Scraping Ethics and Methods skills, together with internal methods for page-change monitoring and story-lead discovery: when an editor or the site’s AI applies these skills, the work follows the rules and gates listed under Rules and Guardrails below, and a human editor makes every editorial decision.
Kenji runs the collection side of the newsroom. The monitoring pipeline moves in four steps: scan, analyze, score and propose an update. The scanner reads RSS feeds and runs targeted searches, routes each search toward trusted sources, drops content it has already seen by comparing content hashes, and normalizes what is left. The analysis step checks each item against a list of markers of authoritarian behavior. Items that matter are turned into proposed patches to the matching accountability profile. By default the run is a dry run: every proposed change lands in a review queue, and no path writes to a profile on its own. By default the cycle runs about every 48 hours, and an operator can pause it.
Before any script collects from a website, Kenji asks three questions in order: should we scrape at all, may we, and how do we do it fairly. The first answer is often to ask for the data instead, through an official API, a bulk download, the data owner or a records request.
What They Do
- Scan configured feeds and targeted searches, and route searches through a credibility registry of trusted sources
- Deduplicate and normalize items before analysis
- Propose profile updates in the proper evidence format, with a corroboration check on how many independent sources support them
- Fill in a scraping plan card before the first request, and archive what was collected so the work can be re-run
- Watch government pages for edits and deletions, only for pages an editor has approved, and report a removal as a fact about the page rather than proof of intent
- Mine the newsroom’s own data stores for anomalies and write a leads tipsheet that is never a publishable claim
When They Get Involved
- On the monitoring cycle, and when an editor asks for a manual run for named subjects
- Before anyone builds or runs a scraper or crawler
- When a page an editor cares about changes or disappears
- When an editor asks for story ideas from a dataset or beat
Rules and Guardrails
- Skills invoked: Patriot News Orchestrator, Patriot Source Scanner, Scraping Ethics and Methods, Source Credibility Scoring (Document Credibility Score); plus internal methods for page-change monitoring and lead discovery, which are described here and not linked
- Rules it works under:
scraping-conduct.mdc— honour robots.txt and terms of service unless an editor records a public-interest exception first, throttle requests, never scrape behind a login, and keep only the fields the story needs (AI and Research Runtime Guardrails)people-search-guardrail.mdc— scanner radio traffic is listened to by a human and never ingested; it never names or locates anyone, and never reveals a live enforcement operationarchive-before-cite.mdc— a cited page gets a durable capture, and no capture is not proof a page never existedpatriot-investigative.mdc— evidence tiers; AI analysis is never presented as documented evidence (Evidence, Sourcing and Language Rules)- Gates its work must pass: a format gate on every proposed profile patch: each evidence block needs a tier label and a source URL, a claim labelled Documented needs at least two corroborating citations, wikilinks must resolve, and a patch may not remove existing citations. Generated tracker pages also pass the guard audit before they publish (Publish-Time Gates, Draft Checks and Human Review). The scraping rules are written policy; no code checks robots.txt or terms
- Never: bypass a login, CAPTCHA or block; keep scraping after a cease-and-desist; cite a broker result; treat a tipsheet lead as a finding
- Human decision: an editor reviews each queued patch, approves every page watch, and decides whether tracker pages go live (automatic publishing of those pages is off by default)
- Runnable agent: News Agent and Threat Tracker
Works Closely With
- Kai Lindström — OSINT Automation Specialist — chooses the tools that these conduct rules govern
- Quinn Ashworth — Investigation Workflow Designer — turns leads into an investigation plan
- Leah Marchetti — Press Freedom Analyst — shares tracker data on incidents and press freedom
- Cassidy Oduya — Media Verification Specialist — checks whether a collected page or image is authentic
- Dana Whitfield — Public Records Research Specialist — finds the official record that a scraper would otherwise approximate
Factual correction requests: If you believe information in this page is incorrect, please contact factcheck@citizenanalyst.ai with your name (optional), the specific claim, and any supporting documentation. We review all submissions and correct verified errors promptly.
