long-contextlisted

Send a large document, codebase, log dump, transcript, contract, or multi-file bundle to the reviewer for read-once-and-summarize tasks — extraction, synthesis, fact-finding, question-answering across a corpus too large to read inline. Use when the user says "summarize this big file with the reviewer," "have the reviewer read the whole codebase," "extract from this transcript," "long-context this with the reviewer," "read this whole document," "process this corpus," "what's in this PDF," "what does the contract say about X," "audit this log dump," or any framing that says "read this large input and tell me what's in it." Also offer this proactively when the active primary is about to either (a) sample-read a clearly-oversized file (instead of reading it whole), (b) iterate read-N-lines-then-read-M-more across a single large document, or (c) hold the entire corpus in working context when a one-shot summarize-and-extract would serve better.
sumitake/agent-collab · ★ 0 · Data & Documents · score 63

Install: claude install-skill sumitake/agent-collab

## Unified runtime invocation Resolve the **plugin root** from this loaded file: `SKILL.md` is at `<plugin-root>/skills/<skill-name>/SKILL.md`. Invoke only `python3 "<plugin-root>/coordinator.py"` and send one bounded JSON request on stdin. Before constructing it, read the **Coordinator request schema** in `<plugin-root>/README.md`; never invent fields or route/action pairs. The public coordinator re-observes the active host/model, captures artifact provenance, excludes same-family routes, and verifies the co-packaged native manifest. It runs standalone from the installed plugin. Never discover a provider executable or reconstruct a raw command. Frontmatter `tier` is a routing recommendation, never a coordinator request field. For a review, cross-check, tiebreaker, or fallback over an authored artifact, capture its exact UTF-8 content and observed author model in the optional `artifact` object even when governance is false; never paste it into the prompt as a provenance substitute. # Long-context — read-once-and-summarize on an oversized input When the input is large and the task is **extraction / synthesis / fact-finding** (not writing), the long-context skill offloads the read to the reviewer and asks for structured output back. The cross-family setup matters less here than in the verification skills; this is **collaborative information retrieval**, not independent verification. No verifier-independence rule applies — use whichever model has the right context window and