jamditis/claude-skills-journalism

jamditis/claude‑skills‑journalism – AI‑agent “skills” for newsroom work

What it is – A curated collection of Agent Skills that Claude Code (Anthropic’s LLM‑powered coding assistant) and OpenAI Codex can load to help journalists, researchers, and small newsroom dev teams. Each skill is a self‑contained instruction set, templates, and supporting files that teach the agent how to perform a concrete journalism‑oriented task (e.g., filing FOIA requests, fact‑checking a claim, transcribing an interview, or designing a PDF report).

Why it matters – Modern LLM‑agents are powerful but need domain‑specific prompts and workflows to be reliable. This repo supplies those prompts in a reusable, version‑controlled format, letting teams plug‑in sophisticated newsroom processes without writing the prompts from scratch.

---

Key components

Component	What it does	How you use it
Plugins (e.g., autocontext, dev‑toolkit, journalism‑core)	Register groups of skills inside Claude Code so you can invoke them with slash commands like /journalism-core:fact-check-workflow.	Run the two /plugin commands shown in the README, then restart Claude Code.
Individual skills (e.g., source‑verification, pdf‑playground:proposal)	Single, modular instruction sets that the agent can call directly (/source‑verification).	Copy or symlink the skill’s folder into ~/.claude/skills/<skill‑name>/SKILL.md (or install via the plugin).
Skill catalog (skills‑catalog.yaml)	Machine‑readable index of every skill, its package, lifecycle, and default invocation mode.	Run npm run validate:catalog to verify that your local copy matches the catalog.
Docs site (https://skills.amditis.tech)	Interactive browser, setup guides, and full reference for every skill.	Use it as the primary reference when adding or customizing skills.

---

Highlighted skill groups

Group	Representative skills	Typical newsroom use case
Journalism core	fact-check-workflow, source‑verification, newsroom‑style, crisis‑communications	End‑to‑end investigative reporting: verify claims, check images/video, enforce AP style, and manage breaking‑news protocols.
Research toolkit	web‑archiving, free‑apis‑catalog, digital‑archive	Build reproducible research archives, pull free public‑API data, and preserve web evidence for future reference.
Dev‑toolkit	web‑scraping, director, secure‑auth	Small newsroom tech teams can automate data collection, orchestrate multi‑agent pipelines, and harden internal tools.
PDF / visual design	pdf‑playground, visual‑explainer	Generate branded PDFs (proposals, newsletters) and HTML visual explainers without leaving the LLM environment.
Superjawn	brainstorming, dispatching‑parallel‑agents	Structured research‑first workflow that fans out parallel agents, then reconvenes for review—useful for complex investigative projects.

---

Installation quick‑start (Claude Code)

1. Make sure Claude Code is installed – claude --version.
2. Add the marketplace entry:

   /plugin marketplace add jamditis/claude-skills-journalism
   /plugin install pdf-playground@claude-skills-journalism
   

3. Restart Claude Code.
4. Invoke a skill – e.g., /journalism-core:fact-check-workflow or /dev-toolkit:director.

If you only need a single skill, clone the repo and copy the SKILL.md folder into ~/.claude/skills/ as described in the README.

---

Typical workflow example

1. Create a persistent session (see the “Persistent sessions” guide) so Claude retains context across days.
2. Run /journalism-core:story-pitch – the agent drafts a pitch template based on a brief description.
3. Switch to /research-toolkit:web-archiving – automatically archive the URLs mentioned in the pitch.
4. Fire /journalism-core:fact-check-workflow – the agent extracts claims, gathers evidence, and produces a rating sheet.
5. Finish with /pdf-playground:proposal – turn the fact‑check report into a branded PDF ready for editors.

---

Who should consider using this

• Newsrooms that already experiment with LLM assistants and need repeatable, vetted prompts.
• Freelance journalists or investigators looking for a turnkey toolbox for source verification, FOIA filing, and data‑driven storytelling.
• Research labs or academic groups that want to embed LLM‑driven workflows (e.g., web‑archiving, API catalog lookup) into their pipelines.
• Small dev teams building internal newsroom tools who want pre‑made security and dev‑ops skills (e.g., security‑toolkit:hotpatch).

---

Maintenance & community

• The repository is actively updated (most entries show an Aug 21 2026 timestamp). The skills‑catalog.yaml and the docs site keep the catalog in sync.
• Contributions are encouraged via standard GitHub pull‑request workflow; new skills should follow the Agent Skills specification linked in the README.
• Compatibility matrices for Codex are provided in plans/codex-compatibility-matrix.md for those using OpenAI’s agent platform.

---

Bottom line

jamditis/claude-skills-journalism is a real, production‑oriented library of LLM‑agent prompt modules that turn Claude Code (and, to a lesser extent, OpenAI Codex) into a specialized newsroom assistant. By installing the plugins or individual skills, a journalist can offload repetitive, compliance‑heavy, or data‑intensive tasks to the LLM while retaining control through clearly versioned, auditable skill definitions.