Summary

The "how" skill guides an AI agent through explaining how a subsystem works via a two-phase approach: an explorer phase that maps code structure (entry points, flows, data shapes, state, external calls) with file:line citations, followed by an explainer phase that synthesizes findings into a layered walkthrough for senior engineers. The skill emphasizes evidence-based claims (all statements must cite code) and explains code as-is rather than prescriptively.

Detected Capabilities

Trigger Keywords

Phrases that agents use to match this skill to user intent.

Referenced Domains

External domains referenced in skill content, detected by static analysis.

Use Cases

• Onboard engineers on unfamiliar codebases
• Explain subsystem architecture before making changes
• Document how existing systems work without redesign critique
• Map complex request flows with precise code citations
• Synthesize parallel exploration findings into coherent explanations

Quality Notes

• Skill uses a well-structured two-phase exploration-explanation pattern with clear role separation and parallel-friendly design
• Evidence-based approach (all claims must cite file:line) is excellent and differentiates this from generic documentation
• Explicit output shape section guides the agent on deliverable format
• Rules section clearly bounds scope: explains code as-is, not prescriptively, with good guardrails against speculation
• Reference prompts are well-documented and provide clear templates for subagent coordination
• Size-to-ask guidance (don't provide full walkthrough for routing questions) shows practical thinking about efficiency
• Supporting reference files (explorer-prompt.md, explainer-prompt.md) are comprehensive and self-contained
• MIT license included for clarity
• No error handling guidance for scenarios where code is too large, sparse, or genuinely unmappable