test-subagent-driven-development-integration.sh 11 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357
  1. #!/usr/bin/env bash
  2. # Integration Test: subagent-driven-development workflow
  3. # Actually executes a plan and verifies the new workflow behaviors
  4. set -euo pipefail
  5. SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
  6. source "$SCRIPT_DIR/test-helpers.sh"
  7. echo "========================================"
  8. echo " Integration Test: subagent-driven-development"
  9. echo "========================================"
  10. echo ""
  11. echo "This test executes a real plan using the skill and verifies:"
  12. echo " 1. Plan is read once (not per task)"
  13. echo " 2. Full task text provided to subagents"
  14. echo " 3. Subagents perform self-review"
  15. echo " 4. Spec compliance review before code quality"
  16. echo " 5. Review loops when issues found"
  17. echo " 6. Spec reviewer reads code independently"
  18. echo ""
  19. echo "WARNING: This test may take 10-30 minutes to complete."
  20. echo ""
  21. # Create test project
  22. TEST_PROJECT=$(create_test_project)
  23. echo "Test project: $TEST_PROJECT"
  24. # Trap to cleanup
  25. trap "cleanup_test_project $TEST_PROJECT" EXIT
  26. # Set up minimal Node.js project
  27. cd "$TEST_PROJECT"
  28. cat > package.json <<'EOF'
  29. {
  30. "name": "test-project",
  31. "version": "1.0.0",
  32. "type": "module",
  33. "scripts": {
  34. "test": "node --test"
  35. }
  36. }
  37. EOF
  38. mkdir -p src test docs/plans
  39. # Create a simple implementation plan
  40. cat > docs/plans/implementation-plan.md <<'EOF'
  41. # Test Implementation Plan
  42. This is a minimal plan to test the subagent-driven-development workflow.
  43. ## Task 1: Create Add Function
  44. Create a function that adds two numbers.
  45. **File:** `src/math.js`
  46. **Requirements:**
  47. - Function named `add`
  48. - Takes two parameters: `a` and `b`
  49. - Returns the sum of `a` and `b`
  50. - Export the function
  51. **Implementation:**
  52. ```javascript
  53. export function add(a, b) {
  54. return a + b;
  55. }
  56. ```
  57. **Tests:** Create `test/math.test.js` that verifies:
  58. - `add(2, 3)` returns `5`
  59. - `add(0, 0)` returns `0`
  60. - `add(-1, 1)` returns `0`
  61. **Verification:** `npm test`
  62. ## Task 2: Create Multiply Function
  63. Create a function that multiplies two numbers.
  64. **File:** `src/math.js` (add to existing file)
  65. **Requirements:**
  66. - Function named `multiply`
  67. - Takes two parameters: `a` and `b`
  68. - Returns the product of `a` and `b`
  69. - Export the function
  70. - DO NOT add any extra features (like power, divide, etc.)
  71. **Implementation:**
  72. ```javascript
  73. export function multiply(a, b) {
  74. return a * b;
  75. }
  76. ```
  77. **Tests:** Add to `test/math.test.js`:
  78. - `multiply(2, 3)` returns `6`
  79. - `multiply(0, 5)` returns `0`
  80. - `multiply(-2, 3)` returns `-6`
  81. **Verification:** `npm test`
  82. EOF
  83. # Initialize git repo
  84. git init --quiet
  85. git config user.email "test@test.com"
  86. git config user.name "Test User"
  87. git add .
  88. git commit -m "Initial commit" --quiet
  89. echo ""
  90. echo "Project setup complete. Starting execution..."
  91. echo ""
  92. # Run Claude with subagent-driven-development
  93. # Capture full output to analyze
  94. OUTPUT_FILE="$TEST_PROJECT/claude-output.txt"
  95. # Create prompt file
  96. cat > "$TEST_PROJECT/prompt.txt" <<'EOF'
  97. I want you to execute the implementation plan at docs/plans/implementation-plan.md using the subagent-driven-development skill.
  98. IMPORTANT: Follow the skill exactly. I will be verifying that you:
  99. 1. Read the plan once at the beginning
  100. 2. Provide full task text to subagents (don't make them read files)
  101. 3. Ensure subagents do self-review before reporting
  102. 4. Run spec compliance review before code quality review
  103. 5. Use review loops when issues are found
  104. Begin now. Execute the plan.
  105. EOF
  106. # Note: We use a longer timeout since this is integration testing
  107. # Use --allowed-tools to enable tool usage in headless mode
  108. # IMPORTANT: Run from superpowers directory so local dev skills are available
  109. PROMPT="Change to directory $TEST_PROJECT and then execute the implementation plan at docs/plans/implementation-plan.md using the subagent-driven-development skill.
  110. IMPORTANT: Follow the skill exactly. I will be verifying that you:
  111. 1. Read the plan once at the beginning
  112. 2. Provide full task text to subagents (don't make them read files)
  113. 3. Ensure subagents do self-review before reporting
  114. 4. Run spec compliance review before code quality review
  115. 5. Use review loops when issues are found
  116. Begin now. Execute the plan."
  117. echo "Running Claude (output will be shown below and saved to $OUTPUT_FILE)..."
  118. echo "================================================================================"
  119. cd "$SCRIPT_DIR/../.." && timeout 1800 claude -p "$PROMPT" --allowed-tools=all --add-dir "$TEST_PROJECT" --permission-mode bypassPermissions 2>&1 | tee "$OUTPUT_FILE" || {
  120. echo ""
  121. echo "================================================================================"
  122. echo "EXECUTION FAILED (exit code: $?)"
  123. exit 1
  124. }
  125. echo "================================================================================"
  126. echo ""
  127. echo "Execution complete. Analyzing results..."
  128. echo ""
  129. # Read the output
  130. OUTPUT=$(cat "$OUTPUT_FILE")
  131. # Verification tests
  132. FAILED=0
  133. echo "=== Verification Tests ==="
  134. echo ""
  135. # Test 1: Plan should be read once at the beginning
  136. echo "Test 1: Plan read once at beginning..."
  137. if echo "$OUTPUT" | grep -q "Load Plan\|read.*plan\|extract.*tasks"; then
  138. # Check it's near the beginning (within first 20% of output)
  139. total_lines=$(echo "$OUTPUT" | wc -l)
  140. plan_line=$(echo "$OUTPUT" | grep -n "Load Plan\|read.*plan\|extract.*tasks" | head -1 | cut -d: -f1)
  141. threshold=$((total_lines / 5))
  142. if [ "$plan_line" -lt "$threshold" ]; then
  143. echo " [PASS] Plan read early (line $plan_line of $total_lines)"
  144. else
  145. echo " [FAIL] Plan not read early (line $plan_line of $total_lines)"
  146. FAILED=$((FAILED + 1))
  147. fi
  148. # Should NOT re-read for each task
  149. read_count=$(echo "$OUTPUT" | grep -c "read.*plan" || echo "0")
  150. if [ "$read_count" -le 3 ]; then # Allow some mentions but not per-task
  151. echo " [PASS] Plan not re-read per task ($read_count mentions)"
  152. else
  153. echo " [FAIL] Plan read too many times ($read_count mentions)"
  154. FAILED=$((FAILED + 1))
  155. fi
  156. else
  157. echo " [FAIL] No evidence of plan loading"
  158. FAILED=$((FAILED + 1))
  159. fi
  160. echo ""
  161. # Test 2: Full task text provided to subagents
  162. echo "Test 2: Full task text provided to subagents..."
  163. if echo "$OUTPUT" | grep -q "Task Description.*Requirements\|FULL TEXT"; then
  164. echo " [PASS] Task text appears to be provided in prompts"
  165. else
  166. echo " [FAIL] No evidence of full task text in prompts"
  167. FAILED=$((FAILED + 1))
  168. fi
  169. # Should NOT make subagent read files
  170. if echo "$OUTPUT" | grep -q "Read.*docs/plans/implementation-plan.md" | grep -v "I'm reading\|I read"; then
  171. echo " [FAIL] Subagent was told to read plan file"
  172. FAILED=$((FAILED + 1))
  173. else
  174. echo " [PASS] Subagent not told to read plan file"
  175. fi
  176. echo ""
  177. # Test 3: Subagents do self-review
  178. echo "Test 3: Subagents perform self-review..."
  179. if echo "$OUTPUT" | grep -qi "self-review\|self review\|reviewing my work\|look.*with fresh eyes"; then
  180. echo " [PASS] Self-review mentioned"
  181. # Check for self-review findings
  182. if echo "$OUTPUT" | grep -qi "completeness\|quality\|discipline"; then
  183. echo " [PASS] Self-review checklist items mentioned"
  184. else
  185. echo " [WARN] Self-review checklist items not clearly mentioned"
  186. fi
  187. else
  188. echo " [FAIL] No evidence of self-review"
  189. FAILED=$((FAILED + 1))
  190. fi
  191. echo ""
  192. # Test 4: Spec compliance review before code quality
  193. echo "Test 4: Spec compliance review before code quality..."
  194. spec_line=$(echo "$OUTPUT" | grep -ni "spec.*compliance.*review" | head -1 | cut -d: -f1)
  195. code_line=$(echo "$OUTPUT" | grep -ni "code.*quality.*review" | head -1 | cut -d: -f1)
  196. if [ -n "$spec_line" ] && [ -n "$code_line" ]; then
  197. if [ "$spec_line" -lt "$code_line" ]; then
  198. echo " [PASS] Spec compliance review before code quality (line $spec_line < $code_line)"
  199. else
  200. echo " [FAIL] Code quality before spec compliance (line $code_line < $spec_line)"
  201. FAILED=$((FAILED + 1))
  202. fi
  203. else
  204. if [ -z "$spec_line" ]; then
  205. echo " [FAIL] No spec compliance review found"
  206. FAILED=$((FAILED + 1))
  207. fi
  208. if [ -z "$code_line" ]; then
  209. echo " [WARN] No code quality review found (might be acceptable if spec review caught issues)"
  210. fi
  211. fi
  212. echo ""
  213. # Test 5: Spec reviewer is skeptical and reads code
  214. echo "Test 5: Spec reviewer reads code independently..."
  215. if echo "$OUTPUT" | grep -qi "do not trust.*report\|verify.*independently\|reading.*code\|inspecting.*implementation"; then
  216. echo " [PASS] Spec reviewer reads code independently"
  217. else
  218. echo " [FAIL] No evidence of independent code verification"
  219. FAILED=$((FAILED + 1))
  220. fi
  221. echo ""
  222. # Test 6: Implementation actually works
  223. echo "Test 6: Implementation verification..."
  224. if [ -f "$TEST_PROJECT/src/math.js" ]; then
  225. echo " [PASS] src/math.js created"
  226. if grep -q "export function add" "$TEST_PROJECT/src/math.js"; then
  227. echo " [PASS] add function exists"
  228. else
  229. echo " [FAIL] add function missing"
  230. FAILED=$((FAILED + 1))
  231. fi
  232. if grep -q "export function multiply" "$TEST_PROJECT/src/math.js"; then
  233. echo " [PASS] multiply function exists"
  234. else
  235. echo " [FAIL] multiply function missing"
  236. FAILED=$((FAILED + 1))
  237. fi
  238. else
  239. echo " [FAIL] src/math.js not created"
  240. FAILED=$((FAILED + 1))
  241. fi
  242. if [ -f "$TEST_PROJECT/test/math.test.js" ]; then
  243. echo " [PASS] test/math.test.js created"
  244. else
  245. echo " [FAIL] test/math.test.js not created"
  246. FAILED=$((FAILED + 1))
  247. fi
  248. # Try running tests
  249. if cd "$TEST_PROJECT" && npm test > test-output.txt 2>&1; then
  250. echo " [PASS] Tests pass"
  251. else
  252. echo " [FAIL] Tests failed"
  253. cat test-output.txt
  254. FAILED=$((FAILED + 1))
  255. fi
  256. echo ""
  257. # Test 7: Git commits show proper workflow
  258. echo "Test 7: Git commit history..."
  259. commit_count=$(git -C "$TEST_PROJECT" log --oneline | wc -l)
  260. if [ "$commit_count" -gt 2 ]; then # Initial + at least 2 task commits
  261. echo " [PASS] Multiple commits created ($commit_count total)"
  262. else
  263. echo " [FAIL] Too few commits ($commit_count, expected >2)"
  264. FAILED=$((FAILED + 1))
  265. fi
  266. echo ""
  267. # Test 8: Check for extra features (spec compliance should catch)
  268. echo "Test 8: No extra features added (spec compliance)..."
  269. if grep -q "export function divide\|export function power\|export function subtract" "$TEST_PROJECT/src/math.js" 2>/dev/null; then
  270. echo " [WARN] Extra features found (spec review should have caught this)"
  271. # Not failing on this as it tests reviewer effectiveness
  272. else
  273. echo " [PASS] No extra features added"
  274. fi
  275. echo ""
  276. # Summary
  277. echo "========================================"
  278. echo " Test Summary"
  279. echo "========================================"
  280. echo ""
  281. if [ $FAILED -eq 0 ]; then
  282. echo "STATUS: PASSED"
  283. echo "All verification tests passed!"
  284. echo ""
  285. echo "The subagent-driven-development skill correctly:"
  286. echo " ✓ Reads plan once at start"
  287. echo " ✓ Provides full task text to subagents"
  288. echo " ✓ Enforces self-review"
  289. echo " ✓ Runs spec compliance before code quality"
  290. echo " ✓ Spec reviewer verifies independently"
  291. echo " ✓ Produces working implementation"
  292. exit 0
  293. else
  294. echo "STATUS: FAILED"
  295. echo "Failed $FAILED verification tests"
  296. echo ""
  297. echo "Output saved to: $OUTPUT_FILE"
  298. echo ""
  299. echo "Review the output to see what went wrong."
  300. exit 1
  301. fi