Qstack How

QStack how

Build a working mental model from the code. This skill is read-only. It does not edit files, fix critique findings, commit, push, publish, deploy, or message anyone.

Choose the mode

• Explain is the default. Answer how the system works and where behavior lives.
• Critique runs only when the user asks for architectural problems or improvements. Explain the system first, then critique it.

This skill explains mechanism, placement, and ownership. Label inferred motivation as inference instead of presenting it as historical fact.

Investigate

1. Resolve the subject and scope. For harmless ambiguity, state a reasonable interpretation and proceed. Ask only when different interpretations would materially change the answer.
2. Search for the entry point and read the implementation. Trace callers, callees, state changes, data transformations, types, and boundaries until the path from trigger to effect has no hand-waved step.
3. Check repository instructions, configuration, generated code, tests, and external boundaries that change the flow. Do not infer behavior from names.
4. Record exact paths, symbols, and lines. State any path that could not be traced.

Handle a narrow question directly. For a cross-cutting subsystem, read complex exploration. When the answer needs a full onboarding narrative, read structured explanation.

Present

Lead with the direct answer. Then include only the concepts, flow, file map, and gotchas needed to make it understandable. Use a diagram when it materially clarifies several components or state transitions.

For Critique mode, read architecture critique only after the explanation is grounded in the code. Report findings and tradeoffs; do not apply fixes.

Adapted from Lauren Tan's PStack how at commit 60c641e4fad674784b30abcf9f8915dea39df38d under the MIT License. See third-party notices.