Accountability Profile Verification Skill
Editorial & Assessment

Accountability Profile Verification Skill

Skip to main content
< All Topics
Print

Accountability Profile Verification Skill

Posted
Updated
ByPU Publish
0 out of 5 stars
5 Stars 0%
4 Stars 0%
3 Stars 0%
2 Stars 0%
1 Stars 0%

When to Use This Skill

Invoke this skill whenever you are:

  • Creating a new accountability profile
  • Auditing an existing profile for legal exposure or citation gaps
  • Upgrading a thin profile that currently cites only One6Project or has no primary sources
  • Removing One6Project references and replacing them with verified citations
  • Adding court case docket numbers to conviction/indictment claims
  • Verifying association or criminal activity claims before publication
  • Running the full post-build audit after any batch profile update

Important: For any profile where the subject is not a sitting or former elected official or Senate-confirmed appointee, run patriot-private-citizen-inclusion-gate before this skill. That skill confirms eligibility and generates the required Basis for Inclusion disclosure block. This skill then handles evidence quality, citation format, and profile completeness.

Core Principles

  1. Speech is not a crime. Opinions, political positions, and even deliberately misleading commentary are not criminal conduct. Document what people did (actions, decisions, votes, transactions) separately from what they said. Speech may be documented as indicating sympathy or support for actors engaged in criminal activity — but this is precisely stated, not implied.
  1. Conviction ≠ indictment ≠ allegation ≠ association. Each must be labeled and cited precisely.
  1. Absence of a verified source means the claim cannot be published as fact. Tag it [NEEDS VERIFICATION — date] or remove it.
  1. The defamation test. For any negative claim, ask: “If the subject sued us for defamation, what would our defense be?” The answer must be a specific cited primary source or multi-outlet corroborated journalism.
  1. One6Project is not a source. It is a taxonomy. It may have originally informed which individuals were included in the knowledge base, but it is not a citation that justifies any specific claim. All One6Project references must be removed.

Step-by-Step Verification Protocol

Step 1: Identify All Factual Claims

Read the full profile. Extract every factual claim — name, date, organization, action, legal status, financial figure, association — into a mental (or literal) checklist.

Flag immediately:

  • Any claim about criminal activity (conviction, indictment, investigation)
  • Any claim about criminal association
  • Any claim about financial misconduct
  • Any priority assignment (P0/P1/P2)
  • Any use of prohibited language (traitor, coup plotter, lied, criminal)

Step 2: Check Sources Section

Look at the Sources section (and any inline Source: citations).

For each source ask:

  • Is this a primary source (court record, FEC filing, congressional record, official government document)?
  • Is this credibly-reported journalism (major outlet, named sources)?
  • Is this One6Project? → Remove immediately
  • Is this AI-generated inference, a think-tank taxonomy without primary backing, or anonymous? → Remove or demote to Alleged

If One6Project was the only source, proceed to Step 3 (Tavily research) before deciding whether to retain or remove the profile.

Step 3: Tavily Research Protocol

Use the tavily-search and tavily-research skills for verification.

For criminal convictions / indictments:


Search: "[Person name] convicted [crime] [court] [year]"
Search: "[Person name] indicted [charge] [year]"
Search: "United States v. [Name]" OR "[State] v. [Name]"

Look for:

  • Case name and docket number (e.g., United States v. Peters, 21-CR-117)
  • Court name (U.S. District Court for [District]; [State] District Court)
  • Conviction date and charges
  • Sentencing date and sentence
  • Appeals status

For association claims:


Search: "[Person name] [associated organization] [year]"
Search: "[Person name] January 6 role"
Search: "[Person name] [specific alleged action]"

Look for:

  • Congressional testimony, official reports (J6 Select Committee)
  • Court filings naming the person
  • Documented financial/organizational records
  • Multi-outlet journalism (not a single partisan source)

For financial / conflict-of-interest claims:


Search: "[Person name] FEC filings [year]"
Search: "[Person name] financial disclosure [year]"
Search: "[Person name] [specific financial claim] ProPublica OR OpenSecrets OR CREW"

Step 4: Update Citations to Primary Sources

For each verified claim, update the inline citation to include:

  • Publication name or court
  • Article/document title (or case name)
  • Publication date (exact date, not just year)
  • URL where available

Publication date discipline (mandatory): Every source citation — whether inline Source link:, Fact-check source:, or in the Sources section — must include the exact publication date of the cited article or document. When using Tavily for research, capture the published_date field from every search result and extraction. Vague date references like “2025” or “2024 campaign” are insufficient; use the specific date (e.g., “November 15, 2024” or “March 12, 2025”). If the publication date cannot be determined from Tavily results, note this explicitly with [publication date not available] rather than omitting it silently.

Preferred citation formats:

For journalism:


**Source:** [Outlet name], "[Article title]," [Month Day, Year]. [URL]

Example:


**Source:** The New Yorker, "The Face of Donald Trump's Deceptively Savvy Media Strategy," March 14, 2025. https://www.newyorker.com/...

For court records, the full preferred format is:


**Court record:** *[Case name]*, [Court], Case No. [Docket], [outcome], [date].

Example:


**Court record:** *People v. Peters*, Mesa County District Court (CO), Case No. 21CR117, convicted on 7 of 10 counts, August 12, 2024; sentenced 9 years, October 4, 2024.

If a URL to the CourtListener or PACER docket is available, add it.

Verifying a court record against CourtListener. PU holds a CourtListener API token (COURTLISTENER_API_TOKEN in .env; Authorization: Token , REST v4 at https://www.courtlistener.com/api/rest/v4/) — use it to confirm the case name, court, docket number and disposition date before publishing, rather than trusting a news report’s rendering of the caption. One targeted lookup per claim is the right cost:


/api/rest/v4/search/?type=r&q=<party name>&court=<court id>   # find the docket
/api/rest/v4/dockets/?id=<docket id>                          # confirm caption + docket number

The quota is small and shared: 10 requests/minute, 75/hour, 300/day per account. A bulk re-verification pass over many profiles will hit the daily cap — batch it, cache every response, and spread it across days rather than retrying against HTTP 429. See docs/courtlistener-api.md.

A CourtListener lookup that returns nothing is not evidence the record does not exist: RECAP coverage of PACER is partial, and state courts are unevenly covered. Treat an empty result as unverified, not disproven, and fall back to Step 5.

Why dates matter: Exact publication dates are required for (a) downstream timeline generation via Cool Timelines Pro, (b) staleness detection during periodic audits, and (c) establishing the temporal sequence of events in accountability documentation. A profile with undated sources cannot support automated timeline extraction.

Step 5: Reclassify Unverifiable Claims

If after Tavily research a claim cannot be verified:

  • Criminal conduct, financial misconduct, or criminal association: Either (a) find a Tier 1/2 source and cite it, (b) demote to “Alleged” with a single-source citation, or (c) remove the claim.
  • Priority/severity: If the P0/P1 designation relied on One6Project as the sole basis, reclassify as P2 (network-level) or lower pending independent verification.
  • Association claims: If the association is with an organization’s criminal activity but the individual’s specific role is unclear, use: “Was affiliated with [Organization], which [documented criminal finding]. [Person]’s individual role in [specific conduct] has not been independently documented.”

Step 6: Apply Language Standards

Review all characterizations. Replace prohibited language:

Remove Replace with
“Coup plotter” (without court record) “Named in the J6 Select Committee report as [specific role]” OR “Coordinated with J6 planning as documented in [source]”
“Lied” / “lying” “Made statements about [X] that were contradicted by [documented evidence]”
“Criminal” (without conviction) “Indicted for / charged with” OR “Under investigation for (as of [date])”
“Traitor” “Convicted of seditious conspiracy” (only if true and cited)
“Funded by [foreign government]” (unverified) “Received [documented amount] from [documented source] [citation]”

Step 7: Add or Verify Factcheck Notice

Ensure the following block appears in the profile before the Sources section:


---

> **Factual correction requests:** If you believe information in this profile is incorrect, please contact [factcheck@patriot.university](mailto:factcheck@patriot.university) with your name (optional), the specific claim, and any supporting documentation. We review all submissions and correct verified errors promptly.

Step 8: Update Metadata

  • Update Last Updated: to today’s date
  • Update Profile Status: to reflect current known status
  • If the profile was significantly revised (claims removed, sources added), add a note: Audit note: Profile revised [date] — One6Project references removed; claims reverified against primary sources.

Step 9: Run Patriot Sanity Check

After completing Steps 1–8, apply the patriot-sanity-check skill to the revised profile:

  • Level 1 (factual claims verification)
  • Level 2 (characterization proportionality)
  • Level 5 (hostile-scrutiny stress test)

Report any CRITICAL or HIGH findings and resolve before finalizing.

Handling Thin Profiles (One6Project Was Only Source)

When a profile’s only source is One6Project taxonomy, choose one of these paths based on what Tavily research finds:

Path A: Verifiable conduct found

  • Document the verified conduct with proper citations
  • Remove One6Project reference
  • Retain or adjust priority based on verified evidence

Path B: No negative conduct verifiable — media/journalistic role only

For Fox News correspondents, journalists, and commentators who appear to have been included based on network affiliation alone:

  • Change profile to document verifiable facts only (career, role, public positions)
  • Remove any criminal-adjacent characterizations
  • Set priority to P2 or lower based on documented role
  • Add note: “Included in this knowledge base as a media-network actor. No individual criminal conduct has been independently documented. This profile documents the individual’s role within [Network]’s documented information environment.”

Path C: Insufficient information for any profile

  • If the subject is a minor figure with no independently verifiable accountability-relevant conduct, consider whether the profile should exist at all.
  • If retaining for network-mapping purposes, strip all negative claims not supported by independent sources and mark clearly as “Network mapping — no individual accountability findings.”

Special Cases

Federal Judges (Article III)

  • Federal judges have a higher threshold for inclusion than other public officials — see Section 5C of patriot-accountability-profile-standards.mdc
  • Erroneous rulings, even rulings that benefit the appointing president, are not sufficient for a personal accountability profile
  • The threshold requires documented personal misconduct: ex parte communications, undisclosed financial interests, personal corruption, formal Judicial Council misconduct finding, or impeachment proceedings
  • When a judge’s rulings have structural impact on accountability but the judge does not clear the personal misconduct threshold, use the Judicial Impact Analysis document format (category: legal; see aileen-cannon-judicial-impact.md for the reference document)
  • Run the J1–J5 gate check from Section 5C before proceeding with any profile for a sitting or former Article III judge

Deceased Subjects

  • Apply same evidentiary standards (defamation risk is lower for deceased public figures but still exists for their estates and living family members)
  • Note date of death in profile
  • Mark status as “Deceased”

Pardoned Individuals

  • Document both the original conviction AND the pardon
  • A pardon does not expunge the conviction from historical record — it can still be accurately reported
  • Note pardon date, granting authority, and scope

Media Personalities (Fox News, etc.)

  • Opinion journalism is protected speech
  • Document specific non-journalistic conduct only (e.g., coordination with political actors, false statements of fact made outside journalistic privilege, financial conflicts)
  • “Promoted false claims about the 2020 election” is documentable fact (Dominion v. Fox settled; Fox internal records were public). “Conspired to steal the election” requires court finding.

Output Quality Checklist

Before finalizing any profile, confirm:

  • [ ] Every claim in Documented Actions has a citation
  • [ ] No One6Project references anywhere in the file
  • [ ] Criminal claims use precise language (convicted/indicted/alleged/investigated)
  • [ ] Court cases include case name, court, and docket or citation where available
  • [ ] Priority (P0/P1/P2) is justified by documented evidence, not external taxonomy
  • [ ] Factcheck notice present before Sources section
  • [ ] Last Updated date is current
  • [ ] No prohibited language (traitor, lying, criminal, coup plotter without documentation)
  • [ ] Association claims specify the individual’s documented role, not just network proximity

Relationship to Other Skills and Rules

Reference Relationship
patriot-accountability-profile-standards.mdc The rule this skill implements
patriot-private-citizen-inclusion-gate Runs BEFORE this skill for non-official subjects; confirms eligibility and generates the Basis for Inclusion block
patriot-accountability-profile-standards.mdc §5C Federal Judge Inclusion Gate — run J1–J5 check before any Article III judge profile
patriot-sanity-check Runs AFTER this skill to validate output
fact-checking General-purpose fact-checking methodology
tavily-search / tavily-research Primary research tools for Step 3
public-records-research-specialist Court records, FOIA, official filings
claims-integrity-audit Cross-platform claims audit

Full Skill Chain for Non-Official Profiles


1. patriot-private-citizen-inclusion-gate  ← Eligibility + Basis for Inclusion block
2. accountability-profile-verification      ← THIS SKILL: evidence standards + citations
3. patriot-sanity-check                    ← Final quality and credibility audit

Master Timeline Repository Integration

After verifying a profile, check whether the master timeline repository already has events for this subject. If it does, use those events as a starting point rather than re-extracting:


from lib.timeline_repo import TimelineRepo

repo = TimelineRepo()
if repo.is_extracted("{slug}"):
    events = repo.events_for_slug("{slug}")
    print(f"Found {len(events)} existing timeline events.")
    # Use these as a verification cross-reference:
    # Do the documented events match the profile claims?

If no events exist in the repository, add extracting events to the post-verification checklist:


make timeline-extract SLUG={slug}

This ensures the master timeline is kept current and other skills/agents can access chronological event data for this profile without re-calling Claude.

Last Updated: May 2026 Canonical location: ~/.codex/skills/accountability-profile-verification/SKILL.md

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?
Table of Contents

Related Articles