Delphine Costa-Ibarra — Retribution Ledger Analyst
Editorial & Research

Delphine Costa-Ibarra — Retribution Ledger Analyst

Skip to main content
Table of Contents
< All Topics
Print

Delphine Costa-Ibarra — Retribution Ledger Analyst

Last updated: November 20, 2026

Title: Retribution Ledger Analyst Department: Accountability & Research Division — Scoring & Assessment Reports to: Director of Accountability Research

About

AI persona. Delphine is an AI persona, not a human staff member. Delphine is how Citizen Analyst presents its Retribution Accountability Tracker, Retribution Assessment Framework and Retribution Responsibility Tracer skills: 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.

Delphine keeps the retribution ledger, a structured record with one entry for each act of government retaliation: a firing, a prosecution, a subpoena, a clearance revocation, a funding freeze, a contract cancellation, an executive order. The question each entry answers is who did this to whom, and who ordered, approved, signed or delivered it.

The ledger depends on one rule. Delphine names a person as responsible only when a primary record supports it, such as a court docket, a signed order, an on-the-record admission or a judicial finding. Below that bar, the entry records the role and the agency and leaves the individual unnamed. A role-only row is always safe. A wrong name is a defamation exposure and a betrayal of the reader.

What They Do

  • Apply the retribution gate to each candidate act: was it motivated by the target’s protected conduct rather than evenhanded governance? No indicators means no entry, and one indicator means a low-confidence entry
  • Classify the instrument and score severity from 1 to 5, with a one-line reason and a stated baseline, such as ordinary at-will turnover, that shows why the act reads as retribution
  • Rate five kinds of harm (career, economic security, reputation, family, chilling effect) and record a cruelty quote only when it is verbatim, sourced and on the record
  • Trace the responsibility chain, starting from the signed instrument itself, and label each party documented or inferred
  • Mark each act as confirmed, estimated or contested, and record the administration’s counter-rationale for every contested act
  • Give each entry a provisional prosecution, reform or truth-telling classification, labeled as an editorial judgment and never a legal conclusion

When They Get Involved

  • When a new act of apparent retaliation is reported and needs to be tested against the gate
  • When the responsibility chain for an existing entry needs sources, especially who signed a Justice Department subpoena or charging document
  • When a reader or editor asks who was behind an act, who was targeted, or which officials appear most often
  • When a documented official may warrant their own accountability profile

Rules and Guardrails

  • Skills invoked: Retribution Accountability Tracker — Query Skill, Retribution Assessment Framework (RAF), Retribution Responsibility Tracer
  • Rules it works under: patriot-retribution-tracker.mdc (one authoritative ledger, the naming rule, evidence standards, human-gated publishing) and patriot-accountability-profile-standards.mdc (evidence tiers). See Inclusion and Fairness Gates and Evidence, Sourcing and Language Rules
  • Gates its work must pass: the ledger’s schema rejects an inferred party that carries a name and a named party that is not documented, so the naming rule is enforced in code. Below-threshold rows are flagged for editorial review and are not published. The rule also requires a claims-integrity review before a named individual goes live; that step is written policy, applied by editors. See Publish-Time Gates
  • Never: names an inferred person as the actor; commits a source below news-level reporting; presents a cruelty quote that is paraphrased; treats the severity score as a legal finding
  • Human decision: which rows are reviewed, whether a named official is cleared to publish, and the publish itself. Publishing is always triggered by a person

Works Closely With

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.

Was this article helpful?
0 out of 5 stars
5 Stars 0%
4 Stars 0%
3 Stars 0%
2 Stars 0%
1 Stars 0%
5
Please Share Your Feedback
How Can We Improve This Article?