62 Skills // 12 Plugins // 17 Hooks

Skills for
journalism.

A curated collection of agent skills for journalists, researchers, academics, and media professionals. Shared skills support Claude Code and Codex, while client-specific components stay clearly labeled.

v1.0.2 · standalone

Skills that do their homework first.

Superjawn is a research-augmented fork of obra/superpowers. A default-on research phase fires before brainstorming, systematic-debugging, and writing-skills — pulling from the web, your codebase, and authoritative docs before any plan gets drafted. Skip it when you already know the answer; otherwise the homework happens automatically. All 14 skills ported, no upstream dependency.

Learn more

Field notes

Field guide · 2026

The Codex review setup I run.

How I set up OpenAI's Codex for code review, how I use it day to day, and the token math behind routing the heavy work to it. A two-audience field guide: the exact command for engineers, the plain version for everyone.

Read the field guide

Find a skill

Every card is taped with the date that skill or plugin last changed. The tape yellows as it ages: under 3 months 3 to 9 months older than 9 months. Use it to spot which ones to check against their sources, or refresh, after you install them.

01 / Plugins

New

okf-wiki

Scaffold an Open Knowledge Format knowledge base: one-concept-per-file markdown with provenance, a spec, a validator that runs in CI, and session-start hooks that orient Claude before it works. Optional GitHub-wiki bootstrap.

Knowledge graph Provenance Validator

Updated

Self-improving

autocontext

Cross-session knowledge persistence with skill evolution. Lessons accumulate per-skill, and /autocontext:evolve folds them back into skill files.

Skill evolution Hooks Confidence

Updated

Research-augmented

superjawn

Research-augmented fork of obra/superpowers. Default-on research phase fires before brainstorming, systematic-debugging, and writing-skills. Includes all 14 skills with no soft dependencies on the upstream superpowers plugin — standalone replacement.

Research phase Subagents Skip protocol

Updated

Interactive Plugin

pdf-playground

Full Claude Code plugin with 6 document types, live preview, brand customization, and conversational editing. Create proposals, reports, slides, and more.

Templates Preview Branding

Updated

Plugin

pdf-design

Standalone skill for PDF reports and proposals with print-ready CSS patterns and layout guidance.

Print CSS Layouts Brand system

Updated

Plugin

journalism-core

Fifteen core journalism skills: AP-style writing, slop detox, source verification (deepfakes/C2PA), FOIA + NJ OPRA, Brazilian LAI requests, fact-checking, interview prep, pitches, editorial workflow, crisis comms, newsletter publishing, and embedded photo metadata.

AP Style FOIA Verification

Updated

Plugin

research-toolkit

Six skills for research and source preservation: academic writing, paywall bypass via Unpaywall, web archiving, page-change monitoring, AI-enriched archives, and a curated free-API catalog with sunset notes.

Academic Archiving APIs

Updated

Plugin

dev-toolkit

Twelve development skills for small newsroom teams: accessibility, context engineering, Electron, mobile debugging, Python pipelines, test-first bug fixing, ethical scraping, no-build frontend, web UI, and CLAUDE.md maintenance.

Accessibility Pipelines Test-first

Updated

Plugin

project-templates-toolkit

Three skills for starting and closing projects: a CLAUDE.md project-memory writer, a LESSONS.md retrospective writer, and a template-selector decision tree across six project types.

CLAUDE.md LESSONS.md Templates

Updated

Plugin

security-toolkit

Four defensive security skills plus /security-toolkit:hotpatch: OWASP audit checklists, secure auth (hashing, JWT, OAuth, passkeys), API hardening, and npm/bun supply-chain hardening.

OWASP Auth Supply chain

Updated

New plugin

video-toolkit

Four-stage social-video reporting pipeline for public downloads, traceable Whisper transcripts, frame analysis, and cross-platform dashboards.

Transcripts Frames Dashboards

Updated

Plugin

visual-explainer

HTML diagrams, data tables, architecture views, slide decks, and KPI dashboards with journalism, newsroom, and academic design sensibilities.

Diagrams Tables Dashboards

Updated

02 / Core journalism

Featured

source-verification

SIFT method, digital verification, reverse image search, social media account analysis, building verification trails.

Updated

Featured

foia-requests

Public records request drafting, tracking systems, appeals process, state-specific guidance.

Updated

New

brazil-records-requests

Brazilian public records request guidance: choose a portal, draft a focused LAI request, track deadlines, and appeal denials.

Updated

Featured

data-journalism

Data acquisition, cleaning, analysis, visualization, and storytelling for newsrooms.

Updated

Style

newsroom-style

AP Style enforcement, attribution rules, headline formatting, number conventions.

Updated

Interviews

interview-prep

Pre-interview research, question frameworks, recording consent, attribution guidelines.

Updated

Transcription

interview-transcription

Recording workflows, transcription, quote management, source databases.

Updated

Pitching

story-pitch

Pitch templates for daily news, features, investigations, op-eds, and freelance queries.

Updated

Fact-check

fact-check-workflow

Claim extraction, evidence gathering, rating scales, correction protocols.

Updated

Editorial

editorial-workflow

Story assignment tracking, deadline management, editorial calendars, handoff protocols.

Updated

OSINT

social-media-intelligence

Social monitoring, narrative tracking, account analysis, coordination detection, OSINT.

Updated

Photos

photo-metadata

Embed caption, byline, credit, alt text, keywords, and license into a photo's IPTC/EXIF/XMP metadata for a news wire.

Updated

03 / Communications

04 / Design & production

05 / Writing quality

06 / Project documentation

07 / Academic and research

08 / Development

Featured

vibe-coding

AI-assisted software development methodology based on YC best practices.

Updated

Featured

one-way-door

Flag irreversible architectural decisions before you commit. Includes PreToolUse hook for automated enforcement.

Updated

Desktop

electron-dev

Electron + React desktop application development patterns.

Updated

Pipeline

python-pipeline

Data processing pipelines with modular architecture.

Updated

Scraping

web-scraping

Content extraction with anti-bot handling and poison pill detection.

Updated

No Build

zero-build-frontend

Static web apps without build tools, CDN-loaded frameworks.

Updated

Mobile

mobile-debugging

Eruda, vConsole, remote debugging, iOS debugging, console capture.

Updated

Featured

accessibility-compliance

WCAG compliance, alt text, accessible charts, keyboard navigation.

Updated

Featured

web-ui-best-practices

Signs of taste in web UI. 18 principles covering speed, restraint, honest cancellation, and optical alignment.

Updated

Setup

persistent-sessions

Keep Claude Code sessions alive across disconnects with tmux. Setup, key bindings, and scheduler coexistence.

Context

claude-md-updater

Propose scoped CLAUDE.md edits as a diff for approval: durable lessons, infra facts, and workflows, never transient session notes.

Updated

Dev

context-engineering-fundamentals

Keep instructions, evidence, and task state usable during long AI-agent sessions.

Updated

09 / Security

10 / Reference

11 / Hooks

Hooks run automatically at specific workflow events. Most are non-blocking warnings, but three block intentionally: one-way-door-check and enforce-test-first stop until you resolve them, and no-ai-attribution denies a commit, PR, or comment that carries AI authorship credit.

Writing quality

PostToolUse

ap-style-check

Flag AP Style violations in written content.

PostToolUse

ai-slop-detector

Warn about AI-generated patterns.

PostToolUse

accessibility-check

Check alt text, heading structure, link text.

UserPromptSubmit

copywriting-preflight

Interview for intent before drafting a piece or revision.

Verification

PostToolUse

source-attribution-check

Flag unattributed quotes and claims.

PostToolUse

verification-reminder

Prompt to verify facts before including.

PostToolUse

data-methodology-check

Ensure data stories include methodology.

Editorial workflow

PostToolUse

source-diversity-check

Note when sources may lack diversity.

PostToolUse

legal-review-flag

Flag potentially defamatory content.

Stop

pre-publish-checklist

Reminder checklist before completing tasks.

SessionStart

deadline-tracker

Surface upcoming deadlines.

Preservation

PostToolUse

archive-reminder

Remind to archive URLs when citing sources.

Development

PreToolUse

one-way-door-check

Block irreversible architectural decisions until discussed.

UserPromptSubmit

bug-report-detector

Detect bug reports and apply test-first workflow.

PreToolUse

enforce-test-first

Block source edits until test is written.

PreToolUse

pre-commit-review

Surface the staged diff for review before a commit.

PreToolUse

no-ai-attribution

Block AI authorship credit in commits, PR bodies, and comments.

12 / Installation

Choose Claude Code or Codex CLI. Check the installed client with claude --version or codex --version.

Recommended

Install as a plugin

Plugins give you slash commands inside Claude Code. Run these two commands in your terminal:

claude plugin marketplace add https://github.com/jamditis/claude-skills-journalism
claude plugin install pdf-playground@claude-skills-journalism

Then restart Claude Code (close and reopen). See the PDF Playground page for detailed instructions, troubleshooting, and brand setup.

Codex compatible

Install a skill package in Codex

Codex can install the 15 nested journalism-core skills through the existing Claude marketplace metadata.

codex plugin marketplace add jamditis/claude-skills-journalism
codex plugin add journalism-core@claude-skills-journalism

This is a verified legacy-compatible route, not a native Codex package. Claude commands, agents, hooks, and root-level skills are not converted. Use one installation route for each skill identity because Codex does not deduplicate identical skills across install roots.

Install individual skills

Skills load automatically when relevant to your work. Claude Code discovers skills at ~/.claude/skills/<skill-name>/SKILL.md, one level deep. Clone the repo anywhere, then copy or symlink the skills you want:

git clone https://github.com/jamditis/claude-skills-journalism.git ~/projects/claude-skills-journalism
cd ~/projects/claude-skills-journalism
mkdir -p ~/.claude/skills
cp -r journalism-core/skills/source-verification ~/.claude/skills/
# or: ln -sfn "$PWD/journalism-core/skills/source-verification" ~/.claude/skills/source-verification

For Codex, the standards-based skills CLI can install one skill into ~/.agents/skills:

npx skills@latest add jamditis/claude-skills-journalism --skill fact-check-workflow --agent codex --copy -g -y

Do not clone the whole repository into a single skill directory. That nests each SKILL.md too deeply for the client to discover.