exploration-strategylisted

# Exploration Strategy > Status: v0.1 baseline draft — derived from the case catalog and axis definitions; awaiting trace evidence from Fable 5.1 case runs. ## When to apply - You are dropped into a repository you have never seen and asked to change or fix something in it. - A task description references symbols, modules, or behaviors you cannot yet locate in the code. - You are about to edit code whose callers, config sources, or side effects you have not confirmed. - A bug report describes a symptom but you do not yet know which layer produces it. - Your context was compacted or the session resumed, and your prior map of the codebase may be stale. - You catch yourself editing based on what a file "probably" contains rather than what you read. ## Core loop 1. **Restate the task as a location question.** Before any file read, write one sentence: "To do X, I need to find where Y happens and what touches it." This defines the exploration target; every subsequent read is judged against it. If you cannot phrase the target, the task itself is underspecified — resolve that first. 2. **Breadth-first structural scan (hard cap: ~10 tool calls).** List the top-level directory tree. Read the manifest (`package.json` / `Cargo.toml` / `go.mod` / `pyproject.toml` / `Makefile`) and the README headings only. Output for yourself: language, build system, test command, and 3-6 candidate directories relevant to the task. Do not open individual source files yet. 3.