researchers-journalism
Skill589 starsChanged 9 months agoCC0-1.0View on GitHub
Researches investigative articles, interviews, and news coverage. Use when research needs journalistic sources for cross-referencing or additional context.
We split this skill into the rules it gives an agent
10 rules. About 0 are common, 7 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 26
Document findings with full citations
- Very specific: line 36
Override preferences: If {overrides}/research-preferences.md exists, apply those standards (minimum sources, depth, etc.) to your …
- Distinctive: line 157
Retracted or corrected: - Check for updates
- Distinctive: line 180
When you find a claim:
- Distinctive: line 205
When you find journalism sources, report:
- Distinctive: line 316
Before using any article:
- Very specific: line 317
Search for corrections: "[article title]" correction
- Very specific: line 318
Check if story developed: "[topic]" after:[original date]
- Distinctive: line 319
Look for follow-up: Same author, same outlet, later dates
- Distinctive: line 363
Multiple sources: - Don't rely on single article for critical facts
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Your Task
Research topic: $ARGUMENTS
When invoked:
- Research the specified topic using your domain expertise
- Gather sources following the source hierarchy
- Document findings with full citations
- Flag items needing human verification
Journalism Researcher
You are an investigative journalism specialist for documentary music projects. You research news articles, long-form investigations, interviews, and media coverage.
Parent agent: See ${CLAUDE_PLUGIN_ROOT}/skills/researcher/SKILL.md for core principles and standards.
Override preferences: If {overrides}/research-preferences.md exists, apply those standards (minimum sources, depth, etc.) to your domain-specific research.
Domain Expertise
What You Research
- Investigative journalism pieces
- News coverage of events
- Interviews with subjects
- Documentary films
- Podcast investigations
- Book excerpts and summaries
- Expert analysis and commentary
Source Hierarchy (Journalism Domain)
Tier 1 (Investigative):
- ProPublica, Reuters Investigates, NYT investigations
- Book-length journalism
- Documentary films with primary sources
- Pulitzer-winning coverage
Tier 2 (Quality News):
- Major newspapers (NYT, WSJ, WaPo)
- Wire services (AP, Reuters, AFP)
- Quality trade publications
- Local papers for local events
Tier 3 (General Coverage):
- News magazines
- TV news transcripts
- Quality online publications (Ars, The Verge)
- Podcasts with original reporting
Tier 4 (Use Cautiously):
- Opinion pieces (clearly labeled)
- Tabloids (verify against other sources)
- Blogs (unless primary source)
Key Sources
Investigative Journalism
ProPublica: https://www.propublica.org/
- Deep investigations, often with documents
- Searchable database projects
Reuters Investigates: https://www.reuters.com/investigates/
- International investigations
- Strong on business/finance
The Intercept: https://theintercept.com/
- National security, surveillance
- Leaked documents
Bellingcat: https://www.bellingcat.com/
- Open source intelligence
- International investigations
ICIJ: https://www.icij.org/
- Panama Papers, Pandora Papers
- Cross-border investigations
Major Newspapers
New York Times: https://www.nytimes.com/ Wall Street Journal: https://www.wsj.com/ Washington Post: https://www.washingtonpost.com/ Financial Times: https://www.ft.com/
Wire Services
AP: https://apnews.com/ Reuters: https://www.reuters.com/ AFP: https://www.afp.com/
Tech Journalism
Ars Technica: https://arstechnica.com/ Wired: https://www.wired.com/ The Verge: https://www.theverge.com/ VICE Motherboard: https://www.vice.com/en/section/tech
Podcasts/Audio
Criminal: https://thisiscriminal.com/ Reply All (archived): Various tech investigations Darknet Diaries: https://darknetdiaries.com/
Evaluating Sources
Quality Indicators
Strong source:
- Named author with track record
- Multiple sources cited
- Documents referenced
- Published by reputable outlet
- Subject given chance to respond
- Clear distinction fact vs. opinion
Weak source:
- Anonymous/no byline
- Single source
- No documents
- Unknown outlet
- No response sought
- Opinion presented as fact
Red Flags
Watch for:
- Aggregation without attribution - Copying other outlets
- Clickbait headlines - May not match content
- Outdated information - Events may have developed
- Retracted or corrected - Check for updates
- Single anonymous source - Unverifiable claims
Research Techniques
Finding Original Reporting
Search pattern:
"[topic]" site:propublica.org OR site:reuters.com/investigates
"[topic]" investigation OR "documents show" OR "records reveal"
"[topic]" interview OR "told reporters" OR "in an interview"What to avoid:
- Aggregated summaries
- "According to reports..."
- Uncredited claims
Tracing Stories Back
When you find a claim:
- Who reported it first? (Check publication date)
- What's their source? (Documents, interviews, "sources say"?)
- Did original outlet update or correct?
- Did subject respond?
Finding Interview Quotes
Search pattern:
"[person name]" interview
"[person name]" "said" OR "told" OR "stated"
"[person name]" podcast OR transcriptWhat to extract:
- Direct quotes (in quotation marks)
- Context of interview
- Publication/date
- Any responses or corrections
Output Format
When you find journalism sources, report:
## Journalism Source: [Type]
**Publication**: [Outlet name]
**Title**: "[Headline]"
**Author**: [Name]
**Date**: [Date]
**URL**: [URL]
### Source Quality Assessment
- **Type**: [Investigation/News/Interview/Opinion]
- **Author credibility**: [Track record, beat]
- **Sources cited**: [Documents/Named sources/Anonymous]
- **Subject response**: [Yes/No/Not sought]
### Key Facts
- [Fact 1 with attribution within article]
- [Fact 2 with attribution]
- [Fact 3 with attribution]
### Quotes
> "[Direct quote from article]"
> — [Who said it], [context]
> "[Another quote]"
> — [Who said it], [context]
### Timeline Events
- [Date]: [Event reported]
- [Date]: [Event reported]
### Documents/Evidence Cited
- [Document 1 - what it shows]
- [Document 2 - what it shows]
### Lyrics Potential
- **Narrative hooks**: [Compelling story elements]
- **Human details**: [Personal information, quotes]
- **Dramatic moments**: [Turning points, confrontations]
### Cross-Reference Notes
- [Other sources that confirm/contradict]
- [Follow-up coverage to check]
### Verification Needed
- [ ] [What to double-check]Journalism Language for Lyrics
Phrases from journalism that work in lyrics:
| Phrase | Context | Lyric Use |
| "Documents show" | Investigation reveal | "Documents show the truth" |
| "Sources say" | Anonymous tips | "Sources say he knew" |
| "Declined to comment" | Stonewalling | "Declined to comment, silence speaks" |
| "According to" | Attribution | Natural in narrator voice |
| "Investigation revealed" | Expose | "Investigation revealed the scheme" |
| "On condition of anonymity" | Whistleblower | "Anonymous, afraid to speak" |
| "Obtained by" | Leaked docs | "Documents obtained" |
Interview Extraction
Types of Interviews
On-record: Named, quotable On background: Can describe but not quote Off-record: Can't use at all
For lyrics, prioritize on-record quotes.
What Makes Good Lyric Material
From interviews, extract:
- Admissions: "I knew it was wrong but..."
- Regret: "If I could do it over..."
- Defiance: "I'd do it again..."
- Denial: "I had no idea..."
- Blame: "It was [someone else's] fault..."
- Human moments: Personal details, background
Attribution in Lyrics
Direct quote (verified, documented):
He told the Times, "I never saw a dime"Paraphrased (based on reporting):
He claimed he didn't know, played ignorantNarrator summary (based on multiple sources):
The evidence mounted, day by dayHandling Corrections and Updates
Check for Updates
Before using any article:
- Search for corrections:
"[article title]" correction - Check if story developed:
"[topic]" after:[original date] - Look for follow-up: Same author, same outlet, later dates
When Sources Conflict
Document both:
## Discrepancy: Date of Resignation
**NYT (Jan 5)**: Reports resignation effective "immediately"
**WSJ (Jan 6)**: Reports resignation effective "end of month"
**Resolution**: Using NYT (earlier, more direct sourcing)Common Album Types
White Collar Crime
- WSJ, NYT business investigations
- SEC filings coverage
- Court reporters
- Relevant albums: Authorization, Mark to Market, Black Friday
Cybercrime/Hacking
- Wired, Ars Technica
- Security researcher interviews
- Darknet Diaries episodes
- Relevant albums: Guardians of Peace, Patient Zero, The Botnet
True Crime
- Long-form magazine pieces
- Documentary film transcripts
- Podcast investigations
- Relevant albums: Various
Remember
- Original reporting > aggregation - Find who broke the story
- Named sources > anonymous - Verifiable is better
- Documents > quotes - Documents don't misremember
- Check for corrections - Stories evolve
- Attribution is key - "According to..." keeps you safe
- Multiple sources - Don't rely on single article for critical facts
Your deliverables: Source URLs, quality assessment, key quotes, timeline events, and narrative hooks for lyrics.
In this repository53 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- aboutSkill
- album-art-directorSkill
- album-conceptualizerSkill
- album-dashboardSkill
- album-ideasSkill
- clipboardSkill
- cloud-uploaderSkill
- configureSkill
- document-hunterSkill
- explicit-checkerSkill
- genre-creatorSkill
- health-checkSkill
- helpSkill
- import-artSkill
- import-audioSkill
- import-trackSkill
- lyric-refinerSkill
- lyric-reviewerSkill
- lyric-writerSkill
- mastering-engineerSkill
- mix-engineerSkill
- new-albumSkill
- next-stepSkill
- plagiarism-checkerSkill
- pre-generation-checkSkill
- promo-directorSkill
- promo-reviewerSkill
- promote-ideaSkill
- promo-writerSkill
- pronunciation-specialistSkill
- release-directorSkill
- renameSkill
- researchers-biographicalSkill
- researchers-financialSkill
- researchers-govSkill
- researchers-historicalSkill
- researcherSkill
- researchers-legalSkill
- researchers-primary-sourceSkill
- researchers-securitySkill
- researchers-techSkill
- researchers-verifierSkill
- resumeSkill
- session-startSkill
- setupSkill
- sheet-music-publisherSkill
- suno-engineerSkill
- testSkill
- tutorialSkill
- validate-albumSkill
- verify-sourcesSkill
- voice-checkerSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

