PR Playbooks for Agents That Usually Write Code

What it does

Newsjack is a bundle of plain-Markdown “skills” plus a small Go CLI that turns terminal agents—Claude Code, Codex, Hermes, OpenClaw—into a PR team. It instructs the agent to monitor industry news, check if a story is still fresh, fit-check journalists, generate story angles, roast pitches against an editor’s rubric, fact-check claims, and even render branded press-clip PDFs by driving a local browser. It is essentially a playbook you install once and delegate to whatever AI already runs your shell.

The interesting bit

Instead of building another SaaS dashboard, the project treats public relations as an agent-native workflow: Markdown instructions the agent reads, a lightweight CLI for state and scheduling, and platform adapters so the same skill runs inside Claude.ai or a local harness. The full experience—scheduled monitoring, saved voice fingerprints, and browser-driven page clipping—only unlocks when the agent lives in your terminal.

Key highlights

• Detects newsjacking opportunities by tracking fresh stories, competitor moves, and keyword coverage with LLM filtering.
• Generates angles, critiques pitches, and extracts a voice fingerprint so drafts don’t read like generic AI output.
• Builds targeted journalist lists with beat fit-checking rather than scraped contact dumps.
• Ships as a curl-installed Go CLI that auto-updates from GitHub Releases, or via npm for the same binary and skill bundle.
• Platform support matrix is refreshingly honest: browser chat apps get a degraded, one-shot “Limited Mode” with no memory between sessions.

Caveats

• Always-on monitoring and the Playwright-based press-clip renderer require a local agent; web chat apps cannot schedule tasks or drive a browser.
• Some skills, like live news search and journalist enrichment, do their best work only when connected to external data sources such as the X API or a Medialyst account.
• ChatGPT support is explicitly limited to Business or Enterprise accounts and runs in a degraded, one-shot mode.

Verdict

Founders doing their own PR and small agencies juggling multiple accounts will get the most leverage, but only if they already live inside a local agent harness. If your AI workflow starts and ends in a browser tab, this is more of a preview than a product.