Durable Streams — Forking

What it does

Forking creates a new durable stream that references (not copies) the data from a source stream up to a specified offset. The fork has its own URL, TTL/expiry, and deletion lifecycle while transparently reading inherited data followed by the fork's own appends. This enables branching workflows, stream variants, and safe experimentation without duplicating data.

When to use it

Use this skill when you need to: create a branched copy of a stream for experimentation; preserve a snapshot view at a known offset; isolate writes to a variant stream while reading inherited history; or manage lifecycle/TTL differences between a derived stream and its source. It is also essential when handling soft-deletes where forks keep data accessible.

What's included

• Scripts: none bundled in this SKILL.md (has_scripts=false)
• References: protocol and client source files are listed in the frontmatter (has_references=false in fetch output but sources are referenced)
• Instructions: step-by-step examples for creating forks (default tail or specific offset), reading forks (transparent stitching of inherited + fork data), writing to forks, TTL/expiry patterns, and safe deletion patterns. Also highlights common mistakes and pitfalls (reusing create handles, fabricating offsets, mismatched content-type, handling 410 Gone).

Compatible agents

Client-focused skill targeting JavaScript/TypeScript agent integrations and tooling that can call Durable Streams client libraries (Copilot/Codex/Gemini CLI/Cursor-friendly snippets).