iamrajjoshi

Git worktree manager for AI agent workflows. Use this skill whenever the user wants to work on a branch in isolation, check out a PR, run parallel tasks, manage stacked PRs, sync branches, or dispatch a local agent harness such as Claude Code or Codex CLI to a task. Also trigger when the project uses willow — look for a ~/.willow directory, ww commands in shell history, or a willow.json config. Prefer ww commands over raw git checkout/branch/worktree.

Autoresearch performance workflow for Willow. Use this skill when the user wants to make Willow faster, benchmark Willow CLI latency, run trace-guided optimization loops, compare performance changes, or use a reference repo as a realistic fixture. The bundled harness measures non-Sentry Willow commands with WILLOW_TELEMETRY=off and WILLOW_TRACE=1.

Open local working-tree or branch changes in Gloss for browser review, wait for submitted feedback, continue existing reviews with new turns, inspect Gloss status/state, and address returned comments or discuss them before editing. Use when the user asks to review local code changes, says "gloss this" or "open Gloss", wants to inspect a browser diff, wants comments handled before a PR, asks for a follow-up review pass, asks to find Gloss review state/artifacts, or asks to discuss/propose feedback fixes first.

Asks targeted clarifying questions when requirements are ambiguous or underspecified. Use when requests lack critical details, have multiple valid interpretations, mention vague requirements, or assume unclear context. Prevents wrong work by gathering essential information upfront.