jsturtevant

{what this skill teaches agents}

Guide for using the Rally CLI tool to dispatch AI agents to GitHub issues and PR reviews via git worktrees. Use this when working with Rally dispatches, the dashboard, or managing worktrees.

Workaround for installing Squad SDK from GitHub branches that don't have dist/ committed

Hard-won patterns for building fullscreen terminal UIs with Ink 5 + React. Covers useInput race conditions, fullscreen rendering, sub-view switching, and testing with ink-testing-library.

Pattern for running multiple sub-agents in parallel using git worktrees. Use this when working on multiple issues/PRs simultaneously.

Dual-review PR workflow: Copilot automated code review + Mal manual team review. All comments must be addressed (fix or explain). Out-of-scope work opens GitHub issues with @copilot assignment.

Proven process for technical design phases: resolve blockers upfront, engage stakeholders early, validate against reality before team review, spec test frameworks before implementation, scope features explicitly.

DEPRECATED: These conventions apply to the Squad CLI tool (create-squad), which is a separate project. Rally uses a different dependency stack.

CodeAct via Hyperlight. Use when looping over many files (8+), cross-referencing results from multiple sources, aggregating data across directories, or chaining 5+ dependent tool calls. Collapses N round-trips into one sandboxed Python run via scripts/codeact.py. NOT beneficial for <5 files or simple grep-then-view — direct tool calls have less overhead at small scale. ESPECIALLY valuable when MCP servers are loaded — fewer turns means the MCP tool catalog context is replayed fewer times. Run `scripts/codeact.py --discover` to see available sandbox tools. Trigger: "codeact", "chain tools", "sandbox", "batch", "for each", "hyperlight sandbox", "collapse tool calls", "run in sandbox", "sandbox execution".

CodeAct via Monty. Use when looping over many files (8+), cross-referencing results from multiple sources, aggregating data across directories, or chaining 5+ dependent tool calls. Collapses N round-trips into one sandboxed Python run via scripts/codeact.py. NOT beneficial for <5 files or simple grep-then-view — direct tool calls have less overhead at small scale. ESPECIALLY valuable when MCP servers are loaded — fewer turns means the MCP tool catalog context is replayed fewer times. Run `scripts/codeact.py --discover` to see available sandbox tools. Sub-microsecond startup. Trigger: "codeact", "chain tools", "sandbox", "batch", "for each", "monty codeact", "monty sandbox", "run in monty", "pydantic monty".

Switch codeact to the Hyperlight backend (micro-VM sandbox via WebAssembly, stronger isolation). Use when user wants to switch to hyperlight, needs full Python support, or stronger sandbox isolation.

Switch codeact to the Monty backend (Pydantic Monty — minimal Python interpreter in Rust, sub-microsecond startup). Use when user wants to switch to monty, use monty backend, or needs lightweight sandboxing.

Install or reconfigure codeact. Auto-detects best backend (monty or hyperlight), runs preflight checks, discovers available tools, and writes configuration files. Use when user asks to install, configure, set up, or reconfigure codeact. Supports --global flag for system-wide install.

Execute JavaScript or Python code in a secure Hyperlight VM sandbox with full isolation. Use this when users request running code, performing calculations, testing algorithms, or demonstrating code behavior with actual output.

{what this skill teaches agents}