fix: give QA fix agents full mock test output instead of 10-line snippets

[AhmedTMM]

Summary

• Phase 3 fix agents now get complete mock test output per cloud, not just 10 lines grepped from the log
• Runs bash test/mock.sh {cloud} for each failing cloud and captures full output to a temp file
• Cleans up temp files after all fix agents complete

Previously, the error context was often too truncated for agents to diagnose root causes — they'd get the tail of one assertion failure but miss the setup context or earlier errors that caused the cascade.

Changes

• .claude/skills/setup-agent-team/qa-cycle.sh: Replace per-script 10-line log grep with per-cloud full mock test capture

Test plan

• Verify bash -n passes on modified script
• Confirm mock output dir is created, populated, and cleaned up
• Verify fix agents receive the full test output in their prompt

Extracted from #833.

🤖 Generated with Claude Code

[louisgv]

Security Review

Verdict: APPROVED

Findings

• [LOW] rm -rf on hardcoded /tmp/spawn-qa-mock-output — constant path, standard cleanup pattern
• [LOW] Unquoted $FAILED_CLOUDS in for loop — intentional word splitting, same pattern as existing code
• [LOW] $cloud in file paths — derived from repo directory names, properly quoted, no traversal risk
• [LOW] Full mock output passed to Claude prompt — trusted local test output, no injection concern

Tests

• bash -n: PASS
• bun test: N/A (no .ts changes)
• curl|bash pattern: N/A (internal skill script)
• macOS compat: OK

---

-- security/pr-reviewer-988

[AhmedTMM]

Summary

• Phase 3 fix agents now get complete mock test output per cloud, not just 10 lines grepped from the log
• Runs bash test/mock.sh {cloud} for each failing cloud and captures full output to a temp file
• Cleans up temp files after all fix agents complete

Previously, the error context was often too truncated for agents to diagnose root causes — they'd get the tail of one assertion failure but miss the setup context or earlier errors that caused the cascade.

Changes

• .claude/skills/setup-agent-team/qa-cycle.sh: Replace per-script 10-line log grep with per-cloud full mock test capture

Test plan

• Verify bash -n passes on modified script
• Confirm mock output dir is created, populated, and cleaned up
• Verify fix agents receive the full test output in their prompt

Extracted from #833.

🤖 Generated with Claude Code

[louisgv]

Security Review

Verdict: APPROVED

Findings

• [LOW] rm -rf on hardcoded /tmp/spawn-qa-mock-output — constant path, standard cleanup pattern
• [LOW] Unquoted $FAILED_CLOUDS in for loop — intentional word splitting, same pattern as existing code
• [LOW] $cloud in file paths — derived from repo directory names, properly quoted, no traversal risk
• [LOW] Full mock output passed to Claude prompt — trusted local test output, no injection concern

Tests

• bash -n: PASS
• bun test: N/A (no .ts changes)
• curl|bash pattern: N/A (internal skill script)
• macOS compat: OK

---

-- security/pr-reviewer-988

[AhmedTMM]

Summary

• Phase 3 fix agents now get complete mock test output per cloud, not just 10 lines grepped from the log
• Runs bash test/mock.sh {cloud} for each failing cloud and captures full output to a temp file
• Cleans up temp files after all fix agents complete

Previously, the error context was often too truncated for agents to diagnose root causes — they'd get the tail of one assertion failure but miss the setup context or earlier errors that caused the cascade.

Changes

• .claude/skills/setup-agent-team/qa-cycle.sh: Replace per-script 10-line log grep with per-cloud full mock test capture

Test plan

• Verify bash -n passes on modified script
• Confirm mock output dir is created, populated, and cleaned up
• Verify fix agents receive the full test output in their prompt

Extracted from #833.

🤖 Generated with Claude Code

[louisgv]

Security Review

Verdict: APPROVED

Findings

• [LOW] rm -rf on hardcoded /tmp/spawn-qa-mock-output — constant path, standard cleanup pattern
• [LOW] Unquoted $FAILED_CLOUDS in for loop — intentional word splitting, same pattern as existing code
• [LOW] $cloud in file paths — derived from repo directory names, properly quoted, no traversal risk
• [LOW] Full mock output passed to Claude prompt — trusted local test output, no injection concern

Tests

• bash -n: PASS
• bun test: N/A (no .ts changes)
• curl|bash pattern: N/A (internal skill script)
• macOS compat: OK

---

-- security/pr-reviewer-988

[AhmedTMM]

Summary

• Phase 3 fix agents now get complete mock test output per cloud, not just 10 lines grepped from the log
• Runs bash test/mock.sh {cloud} for each failing cloud and captures full output to a temp file
• Cleans up temp files after all fix agents complete

Previously, the error context was often too truncated for agents to diagnose root causes — they'd get the tail of one assertion failure but miss the setup context or earlier errors that caused the cascade.

Changes

• .claude/skills/setup-agent-team/qa-cycle.sh: Replace per-script 10-line log grep with per-cloud full mock test capture

Test plan

• Verify bash -n passes on modified script
• Confirm mock output dir is created, populated, and cleaned up
• Verify fix agents receive the full test output in their prompt

Extracted from #833.

🤖 Generated with Claude Code

[louisgv]

Security Review

Verdict: APPROVED

Findings

• [LOW] rm -rf on hardcoded /tmp/spawn-qa-mock-output — constant path, standard cleanup pattern
• [LOW] Unquoted $FAILED_CLOUDS in for loop — intentional word splitting, same pattern as existing code
• [LOW] $cloud in file paths — derived from repo directory names, properly quoted, no traversal risk
• [LOW] Full mock output passed to Claude prompt — trusted local test output, no injection concern

Tests

• bash -n: PASS
• bun test: N/A (no .ts changes)
• curl|bash pattern: N/A (internal skill script)
• macOS compat: OK

---

-- security/pr-reviewer-988

[AhmedTMM]

Summary

• Phase 3 fix agents now get complete mock test output per cloud, not just 10 lines grepped from the log
• Runs bash test/mock.sh {cloud} for each failing cloud and captures full output to a temp file
• Cleans up temp files after all fix agents complete

Previously, the error context was often too truncated for agents to diagnose root causes — they'd get the tail of one assertion failure but miss the setup context or earlier errors that caused the cascade.

Changes

• .claude/skills/setup-agent-team/qa-cycle.sh: Replace per-script 10-line log grep with per-cloud full mock test capture

Test plan

• Verify bash -n passes on modified script
• Confirm mock output dir is created, populated, and cleaned up
• Verify fix agents receive the full test output in their prompt

Extracted from #833.

🤖 Generated with Claude Code

[louisgv]

Security Review

Verdict: APPROVED

Findings

• [LOW] rm -rf on hardcoded /tmp/spawn-qa-mock-output — constant path, standard cleanup pattern
• [LOW] Unquoted $FAILED_CLOUDS in for loop — intentional word splitting, same pattern as existing code
• [LOW] $cloud in file paths — derived from repo directory names, properly quoted, no traversal risk
• [LOW] Full mock output passed to Claude prompt — trusted local test output, no injection concern

Tests

• bash -n: PASS
• bun test: N/A (no .ts changes)
• curl|bash pattern: N/A (internal skill script)
• macOS compat: OK

---

-- security/pr-reviewer-988

[AhmedTMM]

Summary

• Phase 3 fix agents now get complete mock test output per cloud, not just 10 lines grepped from the log
• Runs bash test/mock.sh {cloud} for each failing cloud and captures full output to a temp file
• Cleans up temp files after all fix agents complete

Previously, the error context was often too truncated for agents to diagnose root causes — they'd get the tail of one assertion failure but miss the setup context or earlier errors that caused the cascade.

Changes

• .claude/skills/setup-agent-team/qa-cycle.sh: Replace per-script 10-line log grep with per-cloud full mock test capture

Test plan

• Verify bash -n passes on modified script
• Confirm mock output dir is created, populated, and cleaned up
• Verify fix agents receive the full test output in their prompt

Extracted from #833.

🤖 Generated with Claude Code

[louisgv]

Security Review

Verdict: APPROVED

Findings

• [LOW] rm -rf on hardcoded /tmp/spawn-qa-mock-output — constant path, standard cleanup pattern
• [LOW] Unquoted $FAILED_CLOUDS in for loop — intentional word splitting, same pattern as existing code
• [LOW] $cloud in file paths — derived from repo directory names, properly quoted, no traversal risk
• [LOW] Full mock output passed to Claude prompt — trusted local test output, no injection concern

Tests

• bash -n: PASS
• bun test: N/A (no .ts changes)
• curl|bash pattern: N/A (internal skill script)
• macOS compat: OK

---

-- security/pr-reviewer-988

[AhmedTMM]

Summary

• Phase 3 fix agents now get complete mock test output per cloud, not just 10 lines grepped from the log
• Runs bash test/mock.sh {cloud} for each failing cloud and captures full output to a temp file
• Cleans up temp files after all fix agents complete

Previously, the error context was often too truncated for agents to diagnose root causes — they'd get the tail of one assertion failure but miss the setup context or earlier errors that caused the cascade.

Changes

• .claude/skills/setup-agent-team/qa-cycle.sh: Replace per-script 10-line log grep with per-cloud full mock test capture

Test plan

• Verify bash -n passes on modified script
• Confirm mock output dir is created, populated, and cleaned up
• Verify fix agents receive the full test output in their prompt

Extracted from #833.

🤖 Generated with Claude Code

[louisgv]

Security Review

Verdict: APPROVED

Findings

• [LOW] rm -rf on hardcoded /tmp/spawn-qa-mock-output — constant path, standard cleanup pattern
• [LOW] Unquoted $FAILED_CLOUDS in for loop — intentional word splitting, same pattern as existing code
• [LOW] $cloud in file paths — derived from repo directory names, properly quoted, no traversal risk
• [LOW] Full mock output passed to Claude prompt — trusted local test output, no injection concern

Tests

• bash -n: PASS
• bun test: N/A (no .ts changes)
• curl|bash pattern: N/A (internal skill script)
• macOS compat: OK

---

-- security/pr-reviewer-988

[AhmedTMM]

Summary

• Phase 3 fix agents now get complete mock test output per cloud, not just 10 lines grepped from the log
• Runs bash test/mock.sh {cloud} for each failing cloud and captures full output to a temp file
• Cleans up temp files after all fix agents complete

Previously, the error context was often too truncated for agents to diagnose root causes — they'd get the tail of one assertion failure but miss the setup context or earlier errors that caused the cascade.

Changes

• .claude/skills/setup-agent-team/qa-cycle.sh: Replace per-script 10-line log grep with per-cloud full mock test capture

Test plan

• Verify bash -n passes on modified script
• Confirm mock output dir is created, populated, and cleaned up
• Verify fix agents receive the full test output in their prompt

Extracted from #833.

🤖 Generated with Claude Code

[louisgv]

Security Review

Verdict: APPROVED

Findings

• [LOW] rm -rf on hardcoded /tmp/spawn-qa-mock-output — constant path, standard cleanup pattern
• [LOW] Unquoted $FAILED_CLOUDS in for loop — intentional word splitting, same pattern as existing code
• [LOW] $cloud in file paths — derived from repo directory names, properly quoted, no traversal risk
• [LOW] Full mock output passed to Claude prompt — trusted local test output, no injection concern

Tests

• bash -n: PASS
• bun test: N/A (no .ts changes)
• curl|bash pattern: N/A (internal skill script)
• macOS compat: OK

---

-- security/pr-reviewer-988