
Creates a new git worktree in the auto-co-trader project for any purpose — optimization, regression, backtesting, brainstorming, etc. Use this skill when the user wants to CREATE or SET UP a new worktree — phrases like "prepare a new worktree", "set up a worktree", "create a new worktree for <purpose>", "prep a new worktree", "new worktree for autoresearch", "prepare optimization from [strategy]", or "create a worktree using [strategy]". Do NOT use this skill when the user is already in a worktree and wants to start/run/begin a task — that is handled by the relevant program file in the worktree session.
Use when researching an unfamiliar external API before integration to verify feasibility, integration patterns, and limitations
Use when an agent is asked to define, review, or write acceptance criteria for a request or plan. Derives acceptance criteria from the current request context, confirms them with the user, and writes them into the plan file or a standalone acceptance_criteria.md file.
Use when the user asks to commit changes, create a commit, or save work to git
Use when implementing a fix for a GitHub issue that already has an RCA document at docs/rca/issue-{id}.md
Use when loading project context and architecture overview before starting implementation work, with optional focus on a specific area
Use when generating an implementation report after feature completion documenting what was done, divergences from plan, and test results
Use when performing a meta-level analysis of plan adherence after implementation to identify process improvements and suggest CLAUDE.md updates
Use when running comprehensive project validation including tests, type checking, linting, API connectivity checks, and server startup verification
Use when creating a Product Requirements Document from a conversation or feature request through structured context gathering
Use when fixing bugs found during a code review, processing each issue with explanation and test verification
Use when performing a technical pre-commit code review on recently changed files for bugs, security issues, and standards compliance
Use when setting up or reinitializing this project locally for the first time
Use when creating a git worktree for feature isolation with automatic environment setup including .env files and dependency installation
Use when about to build or modify agent tools and needing to load tool docstring patterns and best practices first
Use when a feature is complete and PROGRESS.md needs cleaning up by replacing verbose execution notes with a concise completion summary
Use when an agent executing an implementation plan claims to have finished, to validate that all acceptance criteria were actually met. Locates acceptance criteria from the plan file, acceptance_criteria.md, or the request itself, then investigates the codebase and surfaces a pass/fail verdict per criterion.
Use when investigating a GitHub issue to identify root cause, assess impact, and create a fix strategy document
Use when creating a comprehensive implementation plan for a new feature including codebase analysis, API research, and parallel execution strategy
Use when implementing a feature plan file, choosing between sequential and team-based parallel execution with mandatory validation gates