{"id":53873,"plugin_id":"plugins_6ac84bac80b081918c305a1574fbd186","kind":"skill","collection_source":"plugin_package","comparison_source":null,"observed_at":"2026-10-24T13:55:15.269Z","digest":"4701a38b269fa8002ab580b9c595c6a4316baaa5e1f7c0dd4f806dda89dc2ebd","against":null,"payload":{"description":"Splits work across multiple agents or sessions running at once, keeping their surfaces disjoint so results merge cleanly. Use this when facing several independent tasks with no shared state, when a plan has parallelizable steps, when a broad search or audit would be faster fanned out, or when deciding whether work can safely be split at all.","included_files":[{"relative_path":"references/sources.md","size_in_bytes":827}],"name":"parallel-agent-delivery","skill_md_contents":"---\nname: parallel-agent-delivery\ndescription: Splits work across multiple agents or sessions running at once, keeping their surfaces disjoint so results merge cleanly. Use this when facing several independent tasks with no shared state, when a plan has parallelizable steps, when a broad search or audit would be faster fanned out, or when deciding whether work can safely be split at all.\n---\n\n# Parallel agent delivery\n\nParallelism is a property of the work, not a preference. Establish it before dispatching.\n\n## When it is safe\n\nAll three must hold:\n\n1. **Disjoint write surfaces** — no two agents touch the same file. Not \"unlikely to conflict\";\n   provably disjoint.\n2. **No sequential dependency** — none needs another's output to start.\n3. **Independently verifiable** — each result can be judged on its own.\n\nFail any one and run sequentially. Two agents editing one file produces a merge conflict at best and\na silent lost update at worst.\n\n## Dispatching\n\nEach agent gets a self-contained brief: the goal, its exclusive surface, the context it cannot see\nfor itself, and the shape of the answer expected. Agents share no memory — anything you leave out is\ngone.\n\nAsk for a **structured return**, the same shape from each, so results can be compared rather than\nre-read.\n\n## Read-only fan-out is the easy case\n\nSearching, auditing, and reviewing have no write surface at all, so they parallelize without risk.\nReach for parallelism here first and freely.\n\n## Integrating\n\n- Review each result against its brief before merging any of them.\n- Where two disagree on a shared fact, neither is authoritative — resolve it yourself.\n- One agent returning nothing useful is a normal outcome, not a failure to retry blindly.\n\n## Sources\n\n`references/sources.md` in this skill lists the outside authorities that settle the questions\nhere — what each one is authoritative for, and what you may do with it. Check them before\nanswering on anything they cover, and cite what you used. Most are free to read and not free\nto reproduce; the use note on each is binding.\n\n## Never\n\n- Split work to look faster when the steps are actually sequential.\n- Let two agents write to one surface because \"they probably won't collide.\"\n- Merge a result you have not read.\n"},"changes":[],"summary":"First saved snapshot. No earlier version is available for comparison.","summary_kind":"deterministic","summary_metadata":{}}