How to use the Release Tweet Writer agent skill

What the release-tweets skill makes the agent do

Once the skill is installed, the agent writes launch tweets, follow-up threads and changelog-to-tweet rewrites for a release that already exists. You ask for a release tweet, launch tweet, X announcement, release thread or social copy for a shipped version, and the agent follows the skill's rules for writing it.

The skill is about release copy only. It does not cut, tag or publish the release itself.

• Reads the relevant CHANGELOG.md section or GitHub release notes before writing
• Checks tag, release, npm, appcast or artifact state where it applies to the project
• Writes one launch tweet under 280 characters with a release URL at the end
• Writes follow-up thread replies one at a time on request
• Marks beta and prerelease versions clearly and avoids implying stable promotion
• Drafts by default and does not post to X unless you explicitly ask

Install the release-tweets skill

Download the release-tweets ZIP from this page. The package contains agents/openai.yaml and SKILL.md. SKILL.md holds the instructions the agent reads, so keep it inside the skill folder when you unzip.

You can install the skill in one of three ways. For a personal setup, unzip it into ~/.claude/skills/. For a single project, unzip it into the project's .claude/skills/ folder. You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.

• Personal: unzip into ~/.claude/skills/ so it is available across your projects
• Project only: unzip into .claude/skills/ inside the project folder
• Claude apps: upload the ZIP under Settings → Capabilities → Skills

Trigger the skill with natural requests

You do not need a special command. Describe the job in ordinary words and mention the release. The skill is meant to activate on requests such as a release tweet, a launch tweet, an X announcement, a release thread, a changelog-to-tweet rewrite, or social copy for a shipped version.

Give the agent the material it needs to check. Point it at the changelog section, the release notes or the version you mean. If you only name a version, the agent still has to verify it, so the more precise the reference, the better the copy.

• Write a launch tweet for version 2.3.0 using the changelog section for that version
• Turn the release notes for the stable release into an X announcement
• Draft a release thread for the new workflow features, starting with the generic launch tweet
• Rewrite this changelog entry as a short tweet with a release link
• Next (after agreeing on the launch tweet, to get the next thread reply)

What the agent checks before it writes

The skill tells the agent to verify the release before writing confident copy. It must distinguish Unreleased entries, beta or prerelease builds, stable releases, hotfixes and correction releases. It must not claim a feature shipped just because it appears at the top of the changelog. Where tag, release or package evidence exists, the agent checks it.

When evidence is incomplete, the agent is told to say what is unverified and to soften its wording. Copy leads with user-visible changes such as features, integrations, workflow improvements, install and update reliability, and security fixes, rather than CI, coverage or refactors.

Launch tweet format and thread follow-ups

The standard launch tweet has the product and version on the first line, then 3-4 compact emoji-led feature bullets, a short punchline and the release or changelog URL at the end. The skill asks for one joke at most and for the feature bullets to carry the message. Each tweet must be under 280 characters, and the agent counts the final characters before calling a tweet ready.

Threads come after the launch tweet is agreed. Each follow-up covers one feature or workflow, is usually 160-220 characters, and includes a docs or release URL for that feature when one is available. Replies do not repeat the version when the thread already states it.

• Use the exact version string and channel, for example a beta number
• Keep one feature per follow-up reply
• Trim filler before trimming facts

Beta, hotfix and correction copy

For a beta or prerelease, the copy makes the beta status explicit and phrases the version as beta N, VERSION-beta.N or preview, as appropriate. It avoids anything that suggests the build is stable.

For a hotfix or correction, the copy is direct. It says what went wrong, what is fixed and what the new version is. The skill asks for no jokes in this case unless you request a lighter tone.

Requirements, copying and posting

The skill needs access to the project's CHANGELOG.md or GitHub release notes. Copying to the clipboard uses pbcopy on macOS, and the skill says to report when text has been copied. Posting to X is optional: the skill prefers a local bird or xurl workflow if the author's setup has one, and it asks the agent to verify the posted URL afterwards.

The skill never invents media. If you want an image, it should come from an existing release screenshot or asset, or be requested or generated separately.

• Required: access to CHANGELOG.md or GitHub release notes
• Optional: local bird or xurl command-line tools, only if you want posting from that setup
• Optional: pbcopy on macOS for clipboard copy

Safety notes: scripts and network

The skill does not run scripts and does not use the network, according to its contents. Its only file-based instructions are SKILL.md and agents/openai.yaml. The main safety feature is that it drafts instead of posting. Posting to X requires you to ask for it explicitly.

Review any draft before you publish it, especially version numbers, channel labels and feature claims. The skill reduces the risk of overclaiming, but you remain responsible for what goes out under your account.

• No scripts are run by the skill
• No network use is described in the skill
• Drafts by default; posting needs an explicit request

Tips for better release tweets

The quality of the copy depends on the quality of the release record. A clear changelog section with user-facing wording gives the agent the best material. Clear, specific requests also help: name the version, the channel and the link you want included.

• Name the exact version and whether it is beta, stable or a hotfix
• Point the agent at the changelog section or release notes
• Ask for a generic launch tweet first, then request threads one reply at a time with next
• Ask the agent to count characters if you plan to paste the tweet directly