skillz

Optimize a codebase for agent readability: documentation structure, domain organization, discoverability, and knowledge capture. Use when agents struggle to find relevant code, when tribal knowledge lives outside the repo, when onboarding an agent to an existing codebase, or when restructuring docs for agent workflows. Use this skill whenever the user wants to make their codebase easier for agents to navigate or understand. Part of the harness engineering workflow; start with `harness-audit` for overall assessment.

Perform a structured code review covering security, performance, correctness, and readability. Use when asked to review code, check a PR, or audit a file. For deeper security-focused analysis, see `vuln-research` instead.

Generate a conventional-commit message from staged changes. Use when asked to write a commit message or describe staged changes.

Design context management for coding agents: sub-agent decomposition, context resets, progressive disclosure hierarchies, and artifact handoffs. Use when agents lose coherence on long tasks, when context windows fill up, when you need to decompose complex workflows into isolated subtasks, or when setting up sub-agent patterns. Use this skill whenever the user mentions context windows, sub-agents, context management, agent memory, or long-running agent tasks. Part of the harness engineering workflow; start with `harness-audit` for overall assessment.

Generate novel ideas using cross-domain analogical mapping — forcing connections between your problem and semantically distant domains (biology, physics, social systems) to produce genuinely original solutions. Use when brainstorming creatively, generating novel ideas, exploring unconventional approaches, or when the user asks for "creative ideas", "think outside the box", or "brainstorm with analogies". Part of the creativity workflow — feed results into `idea-garden` or write them up with `vibe-research`.