# Generate posts from a code repository without publishing a secret

> Source: https://tareno.co/agents/use-cases/posts-from-your-code-repository
> Language: en
> Let a coding agent turn release notes, product feeds, test suites and markdown posts into drafts for X, Pinterest and Facebook, with commands run and secrets removed. You approve in Tareno.

## Short answer

**Can a coding agent turn my repository into social media posts?**

Yes. A coding agent reads release notes, a product feed, tests or markdown posts, runs the commands it quotes, writes drafts and saves them through Tareno. The checks that matter are that every command was executed and that no internal name or key reaches the draft. Nothing is published until you approve it.

Software teams, shops with a product feed and blogs built from markdown all keep their content in files that a coding agent can read. That makes the repository the best brief a social post can have: it is accurate, current and structured. It also makes it a risk, because the same files hold internal names, ticket prefixes and sometimes keys. The guides below show the pattern for different networks and different kinds of repositories, and each adds the check that suits the material.

## The same job in four industries

- **Developer tools:** A maintainer announces a CLI release with an upgrade command that the agent ran in a clean directory first.
- **E-commerce:** A shop developer turns new entries in the product feed into pins and posts, leaving price and stock to the product page.
- **Publishing:** A blogger with a static site generates a pin for every new markdown post from its front matter.
- **Open source:** A library maintainer turns tests into a week of standalone tips, each example executed on the current version.

## How it works

1. Point the agent at a precise range: a tag, a date or a folder. Vague scope produces vague posts.
2. Ask it to run whatever it quotes: commands, snippets, URLs. Show the output next to the claim.
3. Have it list every internal name, path and key in the draft and rewrite the text without them.
4. Save drafts only, and keep any automated run limited to reading and drafting, so a person approves each request.

## Guides that put it to work

- [Can Codex announce a CLI release on X with an upgrade command that actually works?](https://tareno.co/agents/codex/x): A CLI release post with an upgrade command that was run.
- [Can Claude Code turn your test suite into a week of tips for X?](https://tareno.co/agents/claude-code/x): A week of standalone tips drawn from the test suite.
- [Can Claude Code generate a Pinterest pin for every new post in your static-site repo?](https://tareno.co/agents/claude-code/pinterest): A pin for every new markdown post, from its front matter.
- [Can Cursor draft Pinterest pins from the product feed in your shop's code?](https://tareno.co/agents/cursor/pinterest): Pins for new products from a product feed.
- [Can Cursor draft Facebook posts from your web shop's repository?](https://tareno.co/agents/cursor/facebook): Page posts for new products and shop features.
- [Can Claude Code write a free-software release post for Mastodon with the right hashtags?](https://tareno.co/agents/claude-code/mastodon): A free-software release post that checks the license first.

## What to watch out for

- Do not put price or stock in a post that will stay visible for months. Link to the page that holds them.
- A command in a public post gets copied. Run it first.
- Automated jobs should never hold scheduling or publishing tools.

## Questions

### Is it safe to give an agent access to my repository?

That depends on the agent and your setup. Tareno only receives the text and media that the agent sends, so the safety question is what you let the agent read, and what you check in each draft before approving a request.

### Can this run automatically on every release?

A job can read the repository and create drafts. Keep scheduling and publishing out of it, so every request waits for your approval in the Tareno dashboard before anything reaches a network.

### How do I avoid leaking internal names?

Ask the agent to list every service name, path, ticket prefix and hostname in the draft and to rewrite it without them. Then read the result yourself, because what is sensitive is a human judgment.

### Which networks suit repository content?

X for short releases and tips, Pinterest for tall cheat sheets and product pins, Bluesky for threads, Mastodon for free-software releases and LinkedIn for case studies. The guides show one workflow for each.

## Related

- [Remote MCP setup](https://tareno.co/docs/mcp)
- [Approval workflows](https://tareno.co/features/approval-workflows)
- [Post scheduling](https://tareno.co/features/post-scheduling)
- [Read the replies to a post and turn them into a decision](https://tareno.co/agents/use-cases/read-replies-and-triage)
- [Check a claim against its source before it becomes a post](https://tareno.co/agents/use-cases/check-facts-before-posting)
- [Publish photos and slides with image descriptions that are checked](https://tareno.co/agents/use-cases/photos-with-image-descriptions)

Approval model: Tareno saves drafts directly. Scheduling, publishing, edits to scheduled posts and deletions become pending actions that a person approves in the Tareno dashboard; no agent can approve them.