Context hygiene skills
Your context is often cluttered before you even type
AI agents work at their best when the context window is lean. Reasoning is sharpest in the first part of the window and degrades well before the hard limit, so every token you load has to earn its place. The smaller the context, the sharper the agent.
Much of that context is spent before you even type: governing docs, skills, MCP servers and auto-memory all load at startup. These two skills tackle the two places where clutter builds up: your setup and your agent's memory.
context-checkup
Audits everything that auto-loads at startup: the governing-doc chain, skill descriptions, MCP servers, memory. It measures each source instead of guessing, ranks the trims by payoff and keeps every change reversible. Nothing is touched without your approval.
memory-doctor
Auto-memory is a hidden side-channel into every session's context: facts pile up there that you never review. This skill drains it block by block, relocating each entry to a home you control (a doc, a skill, a rule) or archiving it. You confirm every action. Memory works best as an inbox, not a filing cabinet: entries land, get reviewed and move on.