arena

Plan oversized work (more than one session can hold) as a shared map of decision tickets on the issue tracker, resolved one at a time until the way forward is…

Advisory mode for programming-language design discussions: concrete syntax examples first, check prior art before judging novelty, flag bad practices bluntly.

Fix root causes only - no hacks, no hardcoding, no workarounds, no TODOs.

Sweep reviewer and bot feedback across all open PRs of a repo or org, apply repo-wide fixes everywhere they apply, and drive every PR green.

Take pasted code-review comments (Path/Line/Comment), locate the code, fix what's clearly correct, and surface disagreements as short questions instead of…

Reduce cognitive complexity to <=15 per function without changing behavior.

Full release ritual for the mux-lang multi-repo project: bump versions, write changelog, sync docs/READMEs, lint clean, then hand exact tag commands to Derek…

Use for interrogate, adversarial review, multi-model review, challenge this, stress test this code, find blind spots, or tear this apart.

Strip AI-writing tells and add human voice.

Run and manage long-lived shell commands in background terminals.

invoke this skill when the user asks you to use subagents

Interview the user relentlessly about the codebase until reaching shared understanding, resolving each branch of the decision tree.