tsudo
UserA library of standalone Agent Skills for Claude Code, Codex, and other agent runtimes. Install one skill, not a framework.
Categories
Indexed Skills (4)
altitude-check
Assess one body of work against its own stated goal and name the single next thing to do, reading the decisions taken since the plan was written so a choice that cut the spine of the plan gets caught while every task still looks healthy. Use this when someone asks to zoom out, where they stand, what to work on next, or whether this is still the right work. Not for ranking a whole board by priority or assessing two separate bodies of work at once.
grill-me
Interview a half-formed idea, plan, or decision in dependency-ordered rounds until it is committable, then close with a pre-mortem that converts failure modes into binary pass/fail criteria. Use this when someone asks to think something through, poke holes in a plan, or stress-test it before building. Not for reviewing finished work, writing the plan itself, or a decision already made.
reframe
Turn a substantial analysis or long conversation into a standalone Markdown report, using the restructuring as a second pass over evidence, assumptions, risks, and recommendations. Use this when someone asks to write this up, document a decision, or make findings durable. Not for short factual answers, status updates, or transcripts.
deslopify
Rewrite prose that reads as machine-generated into plain, direct writing, fixing structure first, then sentence constructions, then vocabulary, because fixing word choice over a default essay skeleton still reads as generated. Use this when someone asks to clean up the voice, make writing sound human, or cut the AI slop before it ships. Not for agent-facing instruction text - skill files, procedures, prompts - which use a flat imperative register on purpose.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.