zeus-test-gen

---

name: zeus-test-gen description: Generate AI-authored platform test flows (android.test.json, chrome.test.json, ios.test.json) from task.json and prd.json using generate-tests.sh. Can be triggered manually or automatically after zeus-plan. Supports --version flag and --platforms filter. argument-hint: [--version vN] [--platforms android,chrome,ios] shell: bash

Zeus Test Gen — AI Test Flow Generation

Generate platform-specific end-to-end test scenarios from Zeus planning artifacts. AI (zeus-tester agent) authors the test cases. Humans do not write scenarios manually.

Preconditions

• .zeus/{version}/task.json exists and has at least one task.
• .zeus/{version}/prd.json exists and has at least one story.
• .zeus/schemas/test-flow.schema.json exists.
• .zeus/{version}/config.json exists.
• generate-tests.sh is executable.
• claude CLI is installed and reachable.

Optional arguments

• --version vN — target Zeus version folder (default: main)
• --platforms android,chrome,ios — subset of platforms to generate (default: all three)

Deterministic workflow

1) Check preconditions

Verify all required files exist. If any is missing, print a clear error and abort.

for f in .zeus/{version}/task.json .zeus/{version}/prd.json .zeus/schemas/test-flow.schema.json; do
  [[ -f "$f" ]] || { echo "❌ Missing: $f"; exit 1; }
done

2) Show generation plan

Print a preview before executing:

Platform(s) to generate : android, chrome, ios
Task source             : .zeus/{version}/task.json  (N tasks)
Story source            : .zeus/{version}/prd.json   (N stories)
Output directory        : .zeus/{version}/tests/

Require user confirmation to proceed (yes/no).

3) Run generate-tests.sh

bash .zeus/scripts/generate-tests.sh \
  --version {version} \
  --platforms {platforms}

Stream output to the user in real time.

4) Validate generated files

For each output file:

jq empty .zeus/{version}/tests/{platform}.test.json && echo "✓ valid JSON"

Report scenario counts per platform.

5) Print coverage report

After generation succeeds, output a coverage table:

Platform  | Scenarios | Tasks covered | Coverage
----------|-----------|---------------|----------
android   |         8 |           5/6 |      83%
chrome    |         6 |           5/6 |      83%
ios       |         7 |           5/6 |      83%

Coverage = scenarios that reference a task_id / total task count.

6) Commit test files

git add .zeus/{version}/tests/
git commit -m "test(zeus): generate {platform} test flows for {version}"

One commit per platform generated.

7) Write AI log

Create .zeus/{version}/ai-logs/{ISO-ts}-test-gen.md with the 3-section contract:

## 决策理由
Which platforms were targeted and why, what task coverage gaps exist.

## 执行摘要
Files written, scenario counts, commit SHAs.

## 目标预期
How test coverage supports the north star metric by catching regressions early.

8) Report summary

Return:

• scenario count per platform,
• task coverage percentage,
• recommended next step: /zeus:execute or run tests with platform toolchain.

Regeneration behavior

• By default, existing {platform}.test.json files are skipped (not overwritten).
• To regenerate, pass --force to the underlying script via:

bash .zeus/scripts/generate-tests.sh --version {version} --platforms {platforms} --force

Agent collaboration

Delegates test content generation to zeus-tester agent (invoked inside generate-tests.sh via claude --print). Use zeus-researcher to inspect task coverage gaps before generation if needed.