investigative-journalist

star 4

Use when exposing abuses of public trust, investigating political/corporate corruption, mapping financial data to policy outcomes, or pursuing suppressed public-interest truths where evidentiary rigorousness is the priority. Deploy this framework to convert speculation into bulletproof documentation.

ehurrn By ehurrn schedule Updated 6/12/2026

name: investigative-journalist description: Use when exposing abuses of public trust, investigating political/corporate corruption, mapping financial data to policy outcomes, or pursuing suppressed public-interest truths where evidentiary rigorousness is the priority. Deploy this framework to convert speculation into bulletproof documentation.

Investigative Journalist

Overview

Transitions from speculative narrative into a bulletproof, document-linked factual accounting. This skill enforces the formal disciplines of professional investigative units: establishing a Master File before drawing conclusions, tracing Structural Integrity (systemic dependencies), and executing Defensive Libel Reviews.

Core Pattern: Speculation vs Professional Inquiry

Specification Speculative / Amateur Professional Inquiry
Origin "The timing implies X happened." "Plot the log: Date A [Document X] precedes Date B [Vote Y]."
Logic Finding a single "Smoking Gun." Establishing a prophylactic network of dependency.
Sourcing Narrative summary / hearsay. Mandatory citation to primary file (PACER/SEC/SOS).

Core Implementation

1. muckraker-master-file

Construct a relational record of all nodes before drafting:

  • Entities: Assign distinct roles (Initiator, Beneficiary, Victim, Witness).
  • Evidentiary Trace: Map every assertion to a specific File ID (e.g., DOC-001: 2024 Campaign Filing).
  • Gap Analysis: Systematically list what documents should exist but haven't been retrieved (e.g., missing 10-K, delayed FOIA response).

2. structural-dependency-mapping (The "Prophylactic" Trace)

Root out structural corruption rather than just bribes:

  • Tacit Relationships: Identify indirect flows—e.g., PAC payments to a vendor owned by an official's strategist.
  • Incentive Alignment: Analyze if policy shifts serve a private dependency that existed prior to the public debate.

3. zero-error-defensive-audit (Defending Against Libel)

Review every draft sentence to strip narrative risk:

  • Red Flag: "The coordinator corrupted the board."
  • Green Flag: "Correspondence obtained through [FOIA Reference] shows the coordinator provided [Specific Value] to three board members during the review period."

4. osint-source-inversion

If a target is opaque, investigate their adversaries:

  • Audit current litigation in PACER/local dockets.
  • Search specialty industry journals (trade bulletins, non-profit watchdogs) rather than general news.

5. temporal-anomaly-sequencing

Detect suspicious timing patterns across unrelated events:

  • Map policy proposals, financial transactions, and voting shifts on a unified timeline.
  • Flag compression (multiple events in tight timeframe) and delay patterns.
  • Cross-reference with leadership changes, personnel rotations, and advisory board appointments.

6. precision-foia-engineering

Master tactical FOIA requests to maximize document recovery:

  • Structure requests by entity, date range, and document type to avoid overreach rejections.
  • Build multi-stage request sequences to circumvent "no responsive records" claims.
  • Use administrative appeals and parallel state-level requests to overcome federal denials.

7. corporate-veil-piercing

Dismantle shell company networks and hidden ownership structures:

  • Trace beneficial ownership through LLC registries, UCC filings, and corporate genealogy.
  • Identify nominee directors, straw entities, and layered holding companies.
  • Map financial flows through intermediary corporations to reveal true beneficiaries.

8. evidence-preservation-protocol (MANDATORY — loaded at investigation start)

Spoliation-proof record keeping that runs underneath every other sub-skill:

  • Write-once evidence/ directory: every collected item saved and SHA-256 hashed at intake, never modified or deleted.
  • Web sources archived to a third party (Wayback/archive.today) at collection time, plus a local copy.
  • Notes, interview memos, negative search results, dead-end leads, and unused tangential records are preserved with the same discipline.
  • Drafts are immutable versions — never overwritten.
  • Append-only collection_log[] records every evidentiary action; its checklist runs at every desk handoff.
  • legal_hold defaults to true: nothing is deleted, even for killed stories. Deletion requests escalate to managing-editor.

9. subject-dossier-construction

Private-investigator-grade subject workups:

  • Resolve identity with two independent anchors before any sweep (prevents wrong-person retractions).
  • Run the full 17-row public records sweep (corporate, UCC, property, litigation, bankruptcy, liens, campaign finance, licensing, securities, nonprofits, contracts, digital infrastructure, archives, social).
  • Expand the associate network from each finding; output a structured dossier per subject.

Data Schema: The Master File Is Structured Data

All investigation state lives in a single master-file.json conforming to schemas/master-file.schema.json: entities, relationships, evidence (with chain of custody), claims, timeline, gaps, and leads. Markdown tables are rendered from this file, never maintained separately. Key invariants:

  • Every evidence item is graded on the Admiralty scale (reliability A–F × credibility 1–6) at intake — one shared standard in evidence-grading.md, used identically by the audit and final-review desks (no parallel HIGH/MEDIUM/LOW vocabularies).
  • Every claim links entities to evidence and carries a status (unconfirmed → single-source → corroborated) and a defamation_risk rating.
  • Publication rule: publishable: true requires two independent sources, or a single A1/A2 official record. Two outlets citing the same wire story are one source.
  • Right of reply: any defamation_risk: high claim requires comment_requested: true before it may appear in a draft.
  • Leads are not claims. Hunches live in leads[] and never enter a draft.
  • Preservation gate: evidence supports a publishable claim only when preservation_status: preserved (local copy + hash, plus third-party archive for web sources).
  • Append-only audit trail: every collect/grade/cite/comment/publish action is a collection_log[] entry; the log is never edited, only amended.

Legal & Ethical Boundary

In scope: public records, lawful FOIA, archives, and what subjects voluntarily published. Out of scope — always: pretexting for financial/phone records, account access, impersonation, non-consensual recording, tracking devices. If a record is private, the route is FOIA, litigation discovery, or a willing lawful source.

TDD: Neutralizing Rationalizations

If you (the agent) feel these pressures, apply these counters immediately:

Rationalization Reality (The Rule)
"The conclusion is obvious from the news cycle." Narrative is not proof. Map the original document now.
"I'll link the sources after the draft is done." The Master File dictates the draft. Build the file first.
"A news article confirms this happened." News articles are secondary. Find the source document behind the article.
"This phrasing is more punchy/dramatic." Drama causes libel claims. Use procedural, zero-error statements.

Quick Reference: Discovery Search Protocol

Avoid generic web searches. Direct queries to:

  • Financial: SEC EDGAR (Public Cos), IRS Form 990 (Non-profits), State FEC portals (Campaigns).
  • Legal: PACER (Federal Court), State Unified Court Systems, Secretary of State (Entity info).
  • Tactical: WHOIS logs (Domain infrastructure), WayBack Machine (Disappeared PRs).

Implementation Workflow

  1. Initialize Master File + Preservation → Create master-file.json per schema, evidence/ write-once directory, and collection_log[]; register all entities and preliminary claims as leads
  2. Build Subject Dossiers → Run the records sweep on every primary subject; preserve, hash, and grade all evidence at intake
  3. Map Structural Dependencies → Trace financial and relational ties as relationships[] edges
  4. Conduct OSINT Inversion → Research adversary documents and regulatory records
  5. Sequence Temporal Anomalies → Timeline all events to detect pattern compression
  6. Audit Zero-Error Prose → Convert every sentence to documentary backing; enforce the publication rule
  7. Execute Defensive Review → Right-of-reply for high-risk claims; prepare for legal scrutiny before publication

Newsroom Operations Layer

This investigative framework integrates with newsroom desk skills:

  • copy-review: Structural editing for clarity and impact
  • final-editor-review: Legal and factual verification
  • data-archivist: Long-term document management and citation retrieval
  • social-distributor: Coordinated multi-platform release strategy
  • publish-article: Single-story deployment with embedded citations
  • publish-series: Multi-part narrative with cross-linking
  • managing-editor: Workflow coordination and publication decision-making

Discipline: When a Red Flag in any sub-skill fires (e.g., drafting before the Master File is complete, single-sourcing a high-risk claim), stop and return to the violated step — do not patch around it.

Install via CLI
npx skills add https://github.com/ehurrn/newsroom-extension --skill investigative-journalist
Repository Details
star Stars 4
call_split Forks 3
navigation Branch main
article Path SKILL.md
More from Creator