Skills that make an agent spend fewer tokens — on its own replies and on the context it loads — without losing signal.
This skill should be used for improving context efficiency: context budgeting, observation masking, prefix or KV-cache strategy, partitioning, token-cost reduction, retrieval scoping, and extending…
Spend fewer tokens on the context an agent loads — tool outputs, retrieved documents, file reads, and conversation history — without losing signal. Covers compact wire formats, body-elision on source…
Cut token waste in an agent's own replies and its delegation prompts — no preamble/postamble, no filler restating the question, answer-first structure, and the "scripts do the sorting, models do the…
Stop dumping verbose tool and command output into the context window. Redirect it to a file, extract only the signal with standard tools, and keep the full output retrievable behind a handle — the…