Skip to main content
T1 starter · marketing / creators: the diamond DAG. One read fans into three rewrites that run concurrently, and a single write merges the bundle. The craft lives in the prompts; the engine handles the choreography.

The job

You wrote the post. Now it needs to be a thread, a LinkedIn version and a newsletter blurb: same ideas, three voices. Sequential chatbot sessions drift; this file keeps one source and rewrites it three ways at once.

The shape

The file

social-repurpose.nika

How it works

1

Fan-out by omission

thread, linkedin and newsletter each depend only on post: no deps between them, so the engine runs all three at the same time.
2

with: keeps the merge readable

bundle binds the three outputs to one-letter names in with:, the boundary every task reference crosses: a ${{ tasks.… }} written straight into args: is refused (NIKA-VAR-021). Short names keep a template that interpolates three outputs readable.
3

One multi-line write

The content: is a YAML block scalar with three interpolations: the bundle file is assembled by the engine, no model needed for glue.

Constructs you just used

Run it

What it needs: A local model (Ollama serving qwen3.5:4b) and the sample launch post.
1

Rehearse it

A mock model answers, the sample files come with it, and nothing is written in your folder.
2

Get your own copy

Save the file above as social-repurpose.nika in a new folder, then download the sample it reads:
3

Check it, then run it

The run writes out/social-bundle.md.Any other model works: pass --model <provider>/<name> with its key, and cap a cloud model with --max-cost-usd.

Make it yours

  • Add a fourth branch (YouTube description, Mastodon, press blurb): it’s three lines and it runs in the same wave.
  • Pin a glossary: put your brand terms in a const: entry and interpolate it into each prompt.
  • Want every language too? Chain this into Localization factory.

Level up · T2 chains

Next tier: typed outputs, in-place file edits and a webhook ping. Release day, automated.