agentleFS
patricksavallepatricksavalle/investigate-journalism-skills/.claude/skills/journalistic-article-review/SKILL.md

journalistic-article-review

Skill8 starsChanged 3 months agoApache-2.0View on GitHub

A standalone orchestrator for reviewing journalistic articles, news reports, investigations, op-eds presented as reporting, newsletters, and media explainers for accuracy, sourcing, framing, omissions, headline-body alignment, quote context, evidentiary support, and public-interest reliability. Use when the user asks to review, audit, fact-check, critique, verify, assess bias in, or evaluate whether a journalistic article is fair, accurate, misleading, well-sourced, or publication-ready. Route underlying scientific papers to peer-review and individual empirical claims to scientific-fact-classification when those layers become load-bearing.

The file

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 10 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈10

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Do not use this as the primary tool for:

    line 16
  2. Distinctive:

    When an article reports on a paper, policy document, court filing, dataset, or leaked material, review both layers: this article as …

    line 23
  3. Very specific:

    If the article's central claim rests on a study, route the study to peer-review.

    line 32
  4. Distinctive:

    Use specialist skills rather than reimplementing them:

    line 36
  5. Distinctive:

    Every load-bearing factual claim made by the review carries a warrant:

    line 75
  6. Distinctive:

    If the original article body cannot be found or inspected, stop the review immediately.

    line 104
  7. Distinctive:

    Do not proceed to Phase 0, Article Map, Sourcing Audit, Evidence Load Test, Findings, or Journalistic Verdict.

    line 105
  8. Distinctive:

    Do not infer the article's claims from headlines, snippets, excerpts, summaries, social posts, commentary, derivative reporting, or memory.

    line 106
  9. Very specific:

    Use fallacy-bias-and-manipulation-analysis for rhetorical patterns that become load-bearing.

    line 216
  10. Distinctive:

    Stop duplicating specialist work.

    line 220
  11. Distinctive:

    If Phase -1 fails, use only this stop output:

    line 250
  12. Distinctive:

    If Phase -1 passes, use the normal review output:

    line 264

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Journalistic Article Review

Review an article as journalism: accuracy, sourcing, evidentiary load, headline fairness, framing, omissions, quote context, right of reply, and whether the article earns the public-interest claim it makes.

Activation

Trigger on requests such as: "review this article", "is this reporting fair?", "fact-check this news story", "audit this investigation", "does the headline match the evidence?", "is this article biased or misleading?", "is this publishable?", "what reporting is missing?".

Do not use this as the primary tool for:

  • A scientific paper, manuscript, preprint, or study itself — use peer-review.
  • A single empirical claim with no article context — use scientific-fact-classification.
  • A target profile, image, domain, account, or location investigation — use osint-research.
  • A contested event where the task is to build alternative hypotheses — use investigative-reasoning.

When an article reports on a paper, policy document, court filing, dataset, or leaked material, review both layers: this article as journalism, and the underlying primary as the relevant specialist object.

Relationship To Peer Review

journalistic-article-review and peer-review are sibling orchestrators:

  • This skill asks whether an article accurately, fairly, and proportionately reports what its evidence can bear.
  • peer-review asks whether a scientific paper itself is methodologically sound, statistically adequate, reproducible, and honestly cited.
  • A weak paper can be reported accurately. A strong paper can be reported misleadingly. Keep those verdicts separate.
  • If the article's central claim rests on a study, route the study to peer-review; then return to this skill to test whether the article overstates, understates, or correctly represents the peer-review result.

Pairs With

Use specialist skills rather than reimplementing them:

NeedRoute to
Pre-search hunch, gut feeling, or article-level anomaly signalintuitive-thinking
Claim status, causal language, evidence strengthscientific-fact-classification
Underlying scientific paper, preprint, or cited studypeer-review
Rhetorical manipulation, loaded framing, fallaciesfallacy-bias-and-manipulation-analysis
Author, outlet, source, funder, company, account, image, domainosint-research
Competing narratives about an eventinvestigative-reasoning
Assumption stack or inherited premisefirst-principles-thinking
Corrections, updates, new evidence after a prior verdictbelief-revision

This skill owns the article-level synthesis: headline-body fit, sourcing architecture, editorial framing, missing context, fairness to targets, and publication-readiness.

When This Skill Is Silent Or Ambiguous

First check whether another project skill owns the missing layer: hunch / gut feeling / anomaly signal -> intuitive-thinking; scientific claim status -> scientific-fact-classification; paper methods/statistics/citations/reproducibility -> peer-review; article framing/reporting accuracy -> journalistic-article-review; source identity/funding/public records -> osint-research; contested events or competing narratives -> investigative-reasoning; definitions, hidden assumptions, or argument bedrock -> first-principles-thinking; fallacies/rhetoric/statistical framing tricks -> fallacy-bias-and-manipulation-analysis; new evidence changing a prior verdict -> belief-revision.

If no skill clearly owns the gap, reason from first principles and explicit warrants. Built-in knowledge may suggest hypotheses, search terms, possible failure modes, or questions to verify, but any empirical premise remains (memory — unverified) until traced. Reasoning may connect warranted premises; it may not manufacture premises.

Research Discipline (CLAUDE.md/AGENTS.md)

This skill is standalone. Apply these rules even if CLAUDE.md / AGENTS.md are not loaded:

  • Rule 0 (original article retrieval gate) — Before Phase 0, fetch and inspect the original article under review, or use the complete article text supplied in-session. If the original article cannot be found or inspected after reasonable retrieval attempts, stop the review. Do not reconstruct the article from search snippets, summaries, quoted fragments, commentary, archive metadata, or secondary reporting.
  • Rule 1 (pre-search hypothesis registration) — Before searching, write the article's apparent thesis, the strongest alternative interpretation, and the main ways the article could be right or misleading.
  • Rule 2 (steelman from primary literature) — For any criticised person, institution, claim, or contested position, fetch its own primary statement or strongest advocate source, not only critics' summaries.
  • Rule 3 (primary before secondary) — If the article characterises a paper, court filing, dataset, speech, report, post, image, or document, fetch that primary item before relying on the article's characterisation.
  • Rule 4 (map institutional networks) — Treat sources sharing ownership, funding, mandate, syndication, political alignment, national alignment, or campaign infrastructure as one node until independent corroboration is shown.
  • Rule 5 (Tier 0 priority for time-sensitive claims) — For fast-moving or historical claims, record publication dates and prefer contemporaneous primary records over later retrospectives.
  • Rule 6 (bias self-audit) — End by answering whether the same verdict would have been reached if the politically or socially expected answer ran the other way.
  • Rule 7 (minimum search volumes) — Quick article check: 5-10 fetched sources. Substantial article review: 10-20. Narrative-captured or geopolitical article: 20-40+ and geographically diversified.
  • Rule 8 (hostility check on sources) — For every cited source, name role, stake, funding/ownership/mandate, and national alignment where relevant.
  • Rule 9 (interactive refinement: user contributions are inputs, not warrants) — User-supplied claims are (user-supplied — unverified) until fetched and verified in-session. Treat them as hypotheses, not evidence.
  • Rule 10 (objective report voice) — Write the report as a standalone verdict on the article. Do not refer to the requester in the report prose.

Warrant Labels

Every load-bearing factual claim made by the review carries a warrant:

LabelMeaning
(traced)Followed the evidence chain to a primary source fetched in this session via WebFetch/WebSearch, or an explicit terminal/API fetch where the browser fetch path is unsuitable. State URL + access date.
(deferred to consensus)Relying on a named social/institutional consensus mechanism (literature body, regulatory body, textbook, official record system). Consensus is not scientific warrant; for scientific claims, treat it only as a political/social prior unless traced to reproduced or replicated evidence.
(deferred, fragile)Deferred to consensus, but scientific-fact-classification Phase 6c failure modes apply — funder capture, ideological capture, prestige cascade, replication crisis, publication bias, or similar. State which.
(memory — unverified)Recalled from training data, not verified this session. Permitted only with this label, and never load-bearing without an explicit "this could be wrong" caveat.
(user-supplied — unverified)Provided during interactive refinement and not verified in-session. Never load-bearing on its own; treat as a hypothesis to test or an input to verify.
(intuition — unwarranted)A gut feeling, anomaly signal, or pattern impression. It may generate hypotheses and search leads. It is never evidence, never load-bearing, and cannot revise, refute, or establish a claim.

The consensus mechanisms most often deferred to here are a regulator, a court record system, an official statistics body, or a methods guideline. State narrative pressure is an additional (deferred, fragile) failure mode for reporting on contested events.

Maintain a Sources & Warrants table whenever external facts appear.

Phase -1 — Original Article Retrieval Gate

This gate is mandatory and precedes hypothesis registration.

Acceptable article access:

  • The article URL is fetched and the article body is inspectable in-session.
  • An archived, cached, syndicated, raw/API, or local-file copy is fetched and clearly matches the original article identity.
  • The complete article text is supplied in-session, with any missing metadata recorded as a limitation.

Retrieval attempts should match the article type and may include the canonical URL, outlet search, archive.org or other web archives, cached copies, syndicated republication, raw/API endpoints for platform-hosted posts, and local files supplied by the requester.

Hard stop:

  • If the original article body cannot be found or inspected, stop the review immediately.
  • Do not proceed to Phase 0, Article Map, Sourcing Audit, Evidence Load Test, Findings, or Journalistic Verdict.
  • Do not infer the article's claims from headlines, snippets, excerpts, summaries, social posts, commentary, derivative reporting, or memory.
  • The only permitted output is a short retrieval-failure note listing what was tried and what exact input would allow the review to proceed.

Phase 0 — Pre-Review Setup

Register before search:

ItemEntry
Article identityTitle, author, outlet, publication date, update date, URL
Article typeNews report / investigation / analysis / explainer / interview / newsletter / opinion presented as reporting
Apparent thesisOne sentence
Headline claimWhat the headline implies, separately from the body
Strongest fair readingThe best version of what the article is trying to establish
Alternative hypothesisHow the same facts could support a different conclusion
Failure modes to testOverstatement, omission, source capture, quote distortion, headline mismatch, causal leap, missing right of reply

Phase 1 — Article Map

Extract the article's load-bearing structure:

ElementWhat to capture
Headline / deckExact wording and implied claim
Nut graf / thesisWhat the story asks the reader to believe
Factual claimsCheckable descriptive assertions
Causal claims"Led to", "caused", "because", "sparked", "resulted in"
Claim layersSplit adjacent but distinct claims, especially existence / detection / association / infectivity / causation / proof-certainty
EvidenceDocuments, datasets, interviews, observations, images, studies
SourcesNamed, anonymous, official, expert, affected party, adversarial
TargetsPersons or institutions criticised or accused
CaveatsLimitations, uncertainty, contrary evidence, corrections
Missing primariesAny cited item not linked, quoted, or inspectable

Claim-splitting rule. Do not let a supported neighbouring claim carry a stronger adjacent claim. Evidence that X exists, can be detected, is associated with Y, or temporally precedes Y does not by itself support "X causes Y." Treat each layer as a separate article claim with its own evidence burden. If the article blends the layers, split them before the Evidence Load Test.

Phase 2 — Sourcing Audit

Assess whether the sourcing architecture can bear the story:

SourceRole in articleNamed / anonymousIndependence nodeStake / CoIRight of replyWarrant

Checks:

  • Anonymous sources: Is anonymity justified? Is the claim factual, interpretive, or accusatory? Is there independent documentary support?
  • Expertise fit: Is the expert qualified for the specific claim, or being used outside lane?
  • Source symmetry: Are criticised parties, affected parties, neutral specialists, and primary documents represented proportionately?
  • Independence: Are multiple quoted sources actually one institutional or campaign node?
  • Right of reply: If the article makes serious allegations, did it seek and fairly represent response?

Phase 3 — Evidence Load Test

For each load-bearing article claim:

Article claimEvidence offeredPrimary-source checkVerdictWarrant

Verdicts:

  • Supported — article claim matches fetched evidence.
  • Overstated — evidence exists but supports a weaker claim.
  • Under-contextualised — evidence is real but missing necessary denominator, timeline, limitation, or alternative explanation.
  • Unsupported — article offers no adequate evidence for the claim.
  • Contradicted — fetched evidence points against the article claim.
  • Unverifiable — claim cannot be checked from available sources; label the limitation.

Route individual empirical claims to scientific-fact-classification when classification matters to the verdict.

Causal Claim Gate

If an article makes a causal claim ("X causes Y", "X led to Y", "Y is due to X", "X is the cause of disease/outcome Y"), the claim may be marked Supported only when the article's evidence, or the primary evidence fetched during review, addresses causal direction.

Minimum gate:

RequirementPass condition
Reverse causationRuled out by design, directly tested, or made implausible by mechanism and timing
TemporalityCause clearly precedes effect at the relevant biological/social timescale
Alternative causesMajor plausible common causes, confounders, selection effects, and passenger-marker explanations addressed
Intervention / challenge / negative-control evidencePresent where ethically and practically possible, or limitation explicitly caveated
Measurement-layer separationDetection, association, infectivity, and causation are not treated as interchangeable

If reverse causation is not addressed, the causal claim is not demonstrated. Downgrade the verdict for that claim to Overstated, Under-contextualised, or Unsupported depending on the remaining evidence. Do not write "causation is supported but reverse causation is a caveat"; unresolved reverse causation means causation has not been shown by that article.

Severity default:

  • Major — reverse causation is unaddressed for a central causal claim, but the article's main verdict also rests on independently supported non-causal claims.
  • Fatal — the headline, rating, or central article verdict depends on a causal claim whose direction is not established.

Phase 4 — Headline, Framing, And Omission

Audit editorial presentation separately from factual accuracy:

DimensionQuestion
Headline-body fitDoes the article body substantiate what the headline implies?
Nut-graf loadDoes the strongest claim appear before sufficient evidence or caveat?
Buried caveatAre limitations placed after the reader has already absorbed the stronger claim?
Quote contextDo quoted fragments preserve the source's meaning in context?
Loaded languageAre adjectives or verbs doing evidentiary work?
Missing denominatorAre rates, baselines, timeframes, or comparison groups omitted?
Relative-vs-absolute effectAre absolute risk, relative risk, baseline rate, timeframe, and population kept separate?
Missing alternativeIs a plausible alternative explanation ignored?
Causal-direction gapDoes a causal claim fail the Causal Claim Gate, especially by leaving reverse causation unresolved?
Claim-layer collapseAre existence, detection, association, infectivity, causation, and proof-certainty blended as if they were one claim?
False balanceAre weak and strong sources given equal weight?
False certaintyIs unsettled evidence written as settled fact?
Temporal framingDoes the article blur what was known then vs. known now?

Use fallacy-bias-and-manipulation-analysis for rhetorical patterns that become load-bearing.

Phase 5 — Specialist Routing

Stop duplicating specialist work; route it:

Article dependencySpecialist check
Scientific paper carries the articleRun peer-review on the paper's methods, citations, stats, reproducibility, and deployment gap
Single empirical or causal claim carries the articleRun scientific-fact-classification
Outlet/source/author identity is materialRun osint-research
Story sits inside contested event narrativeRun investigative-reasoning
Argument relies on hidden premiseRun first-principles-thinking
New correction or evidence appearsRun belief-revision

Return to this skill after routing and ask: did the article accurately represent what the specialist check found?

Phase 6 — Journalistic Verdict

Grade the article, not the underlying world:

VerdictMeaning
Reliable as reportedMain claims are sourced, proportionate, fairly framed, and caveated.
Mostly reliable with caveatsCore is supported, but there are minor overstatements, omissions, or sourcing limitations.
Mixed / requires further reportingSome claims are supported, but important claims remain weak, unverifiable, or under-contextualised.
MisleadingArticle leads readers toward a conclusion stronger or different than the evidence supports.
UnsupportedCentral claims lack adequate evidence.
ContradictedPrimary evidence contradicts central claims.

Severity tags: Fatal, Major, Minor, Optional, Praise.

Output

If Phase -1 fails, use only this stop output:

# Review Stopped: Original Article Not Found

The review cannot proceed because the original article body was not found or inspectable in-session.

## Retrieval Attempts
- [URL/search/archive/local path tried + result]

## Needed To Proceed
- Original article URL, archived copy, local file, or complete article text.

If Phase -1 passes, use the normal review output:

# Journalistic Article Review: [title]

## Summary
- **Article:** [title, outlet, author, date]
- **Original article access:** [canonical URL / archive / local file / supplied full text; access date]
- **Article type:**
- **Verdict:** Reliable as reported / Mostly reliable with caveats / Mixed / Misleading / Unsupported / Contradicted
- **Bottom line:** [one sentence]

## Article Map
[headline claim, thesis, load-bearing claims, sources, targets, missing primaries]

## Sourcing Audit
[source table with roles, independence, stake, right of reply, warrant]

## Evidence Load Test
[claim-by-claim table: article claim / offered evidence / primary-source check / verdict / warrant]

## Headline, Framing, And Omission
[headline-body fit, caveats, quote context, missing context, loaded language]

## Specialist Checks
[which specialist skills were invoked or should be invoked, and what article-level dependency they affect]

## Findings
[Fatal / Major / Minor / Optional / Praise findings, each with quoted article text, named fault or strength, and fix]

## Sources & Warrants
| Review finding | Source | URL | Access date | Publication date | Warrant | Funding / ownership / mandate / alignment |
|---|---|---|---|---|---|---|

## What Would Change This
- Upward:
- Downward:

## Self-Audit
- **Symmetry test:** Would the same verdict have been reached if the politically/socially expected answer ran the other way? Name the specific judgements (the Evidence Load Test verdicts, the source-independence calls, the severity tags) where the verdict is most sensitive to the prior — asserting symmetry flatly, without identifying where it could break, claims the property rather than showing it. If no — explain. If you can't tell — say so.
- **Source-network test:** If the largest institutional/source node were removed, what independent evidence remains?
- **Requester-input test:** Any user-supplied inputs are labelled and non-load-bearing unless traced.

## Limits
- Sources not accessible:
- Claims not checked:
- Specialist review still needed:

Quick Reference

PatternRiskDefault move
Original article body cannot be fetched or inspectedReview would be reconstructed from fragmentsStop; no verdict or article-level findings
Headline says "X proves Y"; body says "may suggest"Headline overreachMajor or Fatal depending on centrality
Three experts from one advocacy networkFalse independenceCollapse to one node and seek independent source
Anonymous official source makes accusationUncheckable authorityRequire document, named corroboration, or strong caveat
Study reported as causation but design is observationalCausal overreachRoute to peer-review / scientific-fact-classification
Causal claim with reverse direction unaddressedDirectionality failureDowngrade wording to association or route to causal specialist check
Relative risk reduction reported without absolute riskMagnitude inflationRecompute or request event rates; route central claims to scientific-fact-classification
Odds/hazard/rate ratio reported as riskMetric substitutionName the metric exactly and avoid risk-language unless conversion is justified
Serious allegation with no right of replyFairness failureMajor unless response was sought and fairly represented
Caveat appears only near the endBuried limitationFlag framing distortion
Quote fragment changes source meaningQuote-context failureFetch original context and compare
Article relies on "experts say" with no names or documentsAuthority fogRequire named sources or downgrade

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.