| name | test-case-tree-design-expert |
| description | Use when the user wants a decision-tree style test case design with inherited setup and per-leaf assertions. |
Model test cases as a decision tree with inherited setup and per-leaf
assertions.
Mode Selection: Check the input. If terse or vague, follow all steps
(1-5). If already comprehensive with explicit decisions, skip to steps 3-5.
-
Brainstorm — Identify all modes, flags, subcommands, decision points,
error states, and happy-path/sad-path outcomes.
-
Clarify — If details are ambiguous, call add-pending-questions with
JSON arguments (each with a question and optional options array). Stop
and wait after calling.
-
Propose Outline — Present a flat list of proposed test cases (name +
one-line description). Call add-pending-questions to get user approval.
-
Build the Decision Tree — Structure as root (feature) → modes (top-level
branches) → decision nodes → leaves (runnable test cases).
-
Produce Output — Create the directory structure:
<feature-slug>-test-cases/
├── README.md # mermaid graph + text tree + test case index
├── SETUP.md # Global setup (inherited)
├── mode-xxx/
│ ├── SETUP.md # Inherits + mode-specific setup
│ └── leaf-a/
│ ├── SETUP.md
│ └── ASSERT.md
...
SETUP.md: Inherited along ancestor chain. Format: Preconditions, Steps,
Context sections, plus optional executable Go code at the end.
ASSERT.md: Per-leaf assertions. Format: Expected, Side Effects, Errors,
Exit Code sections, plus mandatory executable Go Assert function at the end.
README.md: DOT graph, ASCII text tree, test case index table.
Every runnable leaf must have both SETUP.md and ASSERT.md. After writing,
validate with validate_test_case_tree <output-dir>.
