fix(e2e): dynamically calculate DigitalOcean parallel capacity from account limit

[la14-1]

Summary

• _digitalocean_max_parallel() always returned 3, but pre-existing droplets consume quota slots and cause "droplet limit exceeded" errors when tests run 3-in-parallel
• Now queries /v2/account for the actual droplet_limit and subtracts current droplet count to compute available capacity
• Falls back to 3 if the API is unreachable

Root Cause

The QA account has a 3-droplet limit. During this E2E run, a spawn-5v48 droplet (created 2026-03-11) occupied 1 of 3 slots. Batch 1 tried to create 3 droplets (claude, openclaw, zeroclaw) in parallel — the 3rd slot was already taken, causing claude to fail. Same pattern caused opencode to fail in batch 2.

Test plan

• Verify _digitalocean_max_parallel returns 2 when 1 droplet exists on a 3-limit account
• Verify returns 3 when no pre-existing droplets exist

-- qa/e2e-tester

[louisgv]

Security Review

Verdict: APPROVED
Commit: 1dd6c0c

Findings

No critical or high-severity issues found. Two low-severity observations that are acceptable:

• LOW sh/e2e/lib/clouds/digitalocean.sh:367 — Theoretical integer overflow in wc -l output, mitigated by API pagination (200/page max) and practical droplet limits
• LOW sh/e2e/lib/clouds/digitalocean.sh:368 — Arithmetic could produce negative values if API returns inconsistent data, but handled by minimum-of-1 guard on lines 369-370

Security Assessment

✓ Command injection: Clean - no user input, all data from trusted DigitalOcean API
✓ Credential leaks: Clean - uses _do_curl_auth wrapper with temp file (chmod 600), token never in ps output
✓ Injection safety: Clean - fixed regex patterns, quoted arithmetic expansions
✓ Error handling: Robust - all API calls have fallback to safe default (3)
✓ API response validation: Adequate - uses grep with fixed patterns, handles missing/malformed responses
✓ Integer overflow/underflow: Low risk - bounded by API pagination and practical limits

Tests

• bash -n: PASS
• bun test: PASS (1396 tests, 0 failures)
• curl|bash: N/A (library function)
• macOS compat: OK (uses printf, $(( )), no bash 4+ features)

Code Quality

The implementation is well-structured with appropriate fallbacks and error handling. The dynamic capacity calculation is a significant improvement over the hardcoded limit, reducing E2E test failures when pre-existing droplets consume quota slots.

---

-- security/pr-reviewer

[la14-1]

Summary

• _digitalocean_max_parallel() always returned 3, but pre-existing droplets consume quota slots and cause "droplet limit exceeded" errors when tests run 3-in-parallel
• Now queries /v2/account for the actual droplet_limit and subtracts current droplet count to compute available capacity
• Falls back to 3 if the API is unreachable

Root Cause

The QA account has a 3-droplet limit. During this E2E run, a spawn-5v48 droplet (created 2026-03-11) occupied 1 of 3 slots. Batch 1 tried to create 3 droplets (claude, openclaw, zeroclaw) in parallel — the 3rd slot was already taken, causing claude to fail. Same pattern caused opencode to fail in batch 2.

Test plan

• Verify _digitalocean_max_parallel returns 2 when 1 droplet exists on a 3-limit account
• Verify returns 3 when no pre-existing droplets exist

-- qa/e2e-tester

[louisgv]

Security Review

Verdict: APPROVED
Commit: 1dd6c0c

Findings

No critical or high-severity issues found. Two low-severity observations that are acceptable:

• LOW sh/e2e/lib/clouds/digitalocean.sh:367 — Theoretical integer overflow in wc -l output, mitigated by API pagination (200/page max) and practical droplet limits
• LOW sh/e2e/lib/clouds/digitalocean.sh:368 — Arithmetic could produce negative values if API returns inconsistent data, but handled by minimum-of-1 guard on lines 369-370

Security Assessment

✓ Command injection: Clean - no user input, all data from trusted DigitalOcean API
✓ Credential leaks: Clean - uses _do_curl_auth wrapper with temp file (chmod 600), token never in ps output
✓ Injection safety: Clean - fixed regex patterns, quoted arithmetic expansions
✓ Error handling: Robust - all API calls have fallback to safe default (3)
✓ API response validation: Adequate - uses grep with fixed patterns, handles missing/malformed responses
✓ Integer overflow/underflow: Low risk - bounded by API pagination and practical limits

Tests

• bash -n: PASS
• bun test: PASS (1396 tests, 0 failures)
• curl|bash: N/A (library function)
• macOS compat: OK (uses printf, $(( )), no bash 4+ features)

Code Quality

The implementation is well-structured with appropriate fallbacks and error handling. The dynamic capacity calculation is a significant improvement over the hardcoded limit, reducing E2E test failures when pre-existing droplets consume quota slots.

---

-- security/pr-reviewer

[la14-1]

Summary

• _digitalocean_max_parallel() always returned 3, but pre-existing droplets consume quota slots and cause "droplet limit exceeded" errors when tests run 3-in-parallel
• Now queries /v2/account for the actual droplet_limit and subtracts current droplet count to compute available capacity
• Falls back to 3 if the API is unreachable

Root Cause

The QA account has a 3-droplet limit. During this E2E run, a spawn-5v48 droplet (created 2026-03-11) occupied 1 of 3 slots. Batch 1 tried to create 3 droplets (claude, openclaw, zeroclaw) in parallel — the 3rd slot was already taken, causing claude to fail. Same pattern caused opencode to fail in batch 2.

Test plan

• Verify _digitalocean_max_parallel returns 2 when 1 droplet exists on a 3-limit account
• Verify returns 3 when no pre-existing droplets exist

-- qa/e2e-tester

[louisgv]

Security Review

Verdict: APPROVED
Commit: 1dd6c0c

Findings

No critical or high-severity issues found. Two low-severity observations that are acceptable:

• LOW sh/e2e/lib/clouds/digitalocean.sh:367 — Theoretical integer overflow in wc -l output, mitigated by API pagination (200/page max) and practical droplet limits
• LOW sh/e2e/lib/clouds/digitalocean.sh:368 — Arithmetic could produce negative values if API returns inconsistent data, but handled by minimum-of-1 guard on lines 369-370

Security Assessment

✓ Command injection: Clean - no user input, all data from trusted DigitalOcean API
✓ Credential leaks: Clean - uses _do_curl_auth wrapper with temp file (chmod 600), token never in ps output
✓ Injection safety: Clean - fixed regex patterns, quoted arithmetic expansions
✓ Error handling: Robust - all API calls have fallback to safe default (3)
✓ API response validation: Adequate - uses grep with fixed patterns, handles missing/malformed responses
✓ Integer overflow/underflow: Low risk - bounded by API pagination and practical limits

Tests

• bash -n: PASS
• bun test: PASS (1396 tests, 0 failures)
• curl|bash: N/A (library function)
• macOS compat: OK (uses printf, $(( )), no bash 4+ features)

Code Quality

The implementation is well-structured with appropriate fallbacks and error handling. The dynamic capacity calculation is a significant improvement over the hardcoded limit, reducing E2E test failures when pre-existing droplets consume quota slots.

---

-- security/pr-reviewer

[la14-1]

Summary

• _digitalocean_max_parallel() always returned 3, but pre-existing droplets consume quota slots and cause "droplet limit exceeded" errors when tests run 3-in-parallel
• Now queries /v2/account for the actual droplet_limit and subtracts current droplet count to compute available capacity
• Falls back to 3 if the API is unreachable

Root Cause

The QA account has a 3-droplet limit. During this E2E run, a spawn-5v48 droplet (created 2026-03-11) occupied 1 of 3 slots. Batch 1 tried to create 3 droplets (claude, openclaw, zeroclaw) in parallel — the 3rd slot was already taken, causing claude to fail. Same pattern caused opencode to fail in batch 2.

Test plan

• Verify _digitalocean_max_parallel returns 2 when 1 droplet exists on a 3-limit account
• Verify returns 3 when no pre-existing droplets exist

-- qa/e2e-tester

[louisgv]

Security Review

Verdict: APPROVED
Commit: 1dd6c0c

Findings

No critical or high-severity issues found. Two low-severity observations that are acceptable:

• LOW sh/e2e/lib/clouds/digitalocean.sh:367 — Theoretical integer overflow in wc -l output, mitigated by API pagination (200/page max) and practical droplet limits
• LOW sh/e2e/lib/clouds/digitalocean.sh:368 — Arithmetic could produce negative values if API returns inconsistent data, but handled by minimum-of-1 guard on lines 369-370

Security Assessment

✓ Command injection: Clean - no user input, all data from trusted DigitalOcean API
✓ Credential leaks: Clean - uses _do_curl_auth wrapper with temp file (chmod 600), token never in ps output
✓ Injection safety: Clean - fixed regex patterns, quoted arithmetic expansions
✓ Error handling: Robust - all API calls have fallback to safe default (3)
✓ API response validation: Adequate - uses grep with fixed patterns, handles missing/malformed responses
✓ Integer overflow/underflow: Low risk - bounded by API pagination and practical limits

Tests

• bash -n: PASS
• bun test: PASS (1396 tests, 0 failures)
• curl|bash: N/A (library function)
• macOS compat: OK (uses printf, $(( )), no bash 4+ features)

Code Quality

The implementation is well-structured with appropriate fallbacks and error handling. The dynamic capacity calculation is a significant improvement over the hardcoded limit, reducing E2E test failures when pre-existing droplets consume quota slots.

---

-- security/pr-reviewer

[la14-1]

Summary

• _digitalocean_max_parallel() always returned 3, but pre-existing droplets consume quota slots and cause "droplet limit exceeded" errors when tests run 3-in-parallel
• Now queries /v2/account for the actual droplet_limit and subtracts current droplet count to compute available capacity
• Falls back to 3 if the API is unreachable

Root Cause

The QA account has a 3-droplet limit. During this E2E run, a spawn-5v48 droplet (created 2026-03-11) occupied 1 of 3 slots. Batch 1 tried to create 3 droplets (claude, openclaw, zeroclaw) in parallel — the 3rd slot was already taken, causing claude to fail. Same pattern caused opencode to fail in batch 2.

Test plan

• Verify _digitalocean_max_parallel returns 2 when 1 droplet exists on a 3-limit account
• Verify returns 3 when no pre-existing droplets exist

-- qa/e2e-tester

[louisgv]

Security Review

Verdict: APPROVED
Commit: 1dd6c0c

Findings

No critical or high-severity issues found. Two low-severity observations that are acceptable:

• LOW sh/e2e/lib/clouds/digitalocean.sh:367 — Theoretical integer overflow in wc -l output, mitigated by API pagination (200/page max) and practical droplet limits
• LOW sh/e2e/lib/clouds/digitalocean.sh:368 — Arithmetic could produce negative values if API returns inconsistent data, but handled by minimum-of-1 guard on lines 369-370

Security Assessment

✓ Command injection: Clean - no user input, all data from trusted DigitalOcean API
✓ Credential leaks: Clean - uses _do_curl_auth wrapper with temp file (chmod 600), token never in ps output
✓ Injection safety: Clean - fixed regex patterns, quoted arithmetic expansions
✓ Error handling: Robust - all API calls have fallback to safe default (3)
✓ API response validation: Adequate - uses grep with fixed patterns, handles missing/malformed responses
✓ Integer overflow/underflow: Low risk - bounded by API pagination and practical limits

Tests

• bash -n: PASS
• bun test: PASS (1396 tests, 0 failures)
• curl|bash: N/A (library function)
• macOS compat: OK (uses printf, $(( )), no bash 4+ features)

Code Quality

The implementation is well-structured with appropriate fallbacks and error handling. The dynamic capacity calculation is a significant improvement over the hardcoded limit, reducing E2E test failures when pre-existing droplets consume quota slots.

---

-- security/pr-reviewer

[la14-1]

Summary

• _digitalocean_max_parallel() always returned 3, but pre-existing droplets consume quota slots and cause "droplet limit exceeded" errors when tests run 3-in-parallel
• Now queries /v2/account for the actual droplet_limit and subtracts current droplet count to compute available capacity
• Falls back to 3 if the API is unreachable

Root Cause

The QA account has a 3-droplet limit. During this E2E run, a spawn-5v48 droplet (created 2026-03-11) occupied 1 of 3 slots. Batch 1 tried to create 3 droplets (claude, openclaw, zeroclaw) in parallel — the 3rd slot was already taken, causing claude to fail. Same pattern caused opencode to fail in batch 2.

Test plan

• Verify _digitalocean_max_parallel returns 2 when 1 droplet exists on a 3-limit account
• Verify returns 3 when no pre-existing droplets exist

-- qa/e2e-tester

[louisgv]

Security Review

Verdict: APPROVED
Commit: 1dd6c0c

Findings

No critical or high-severity issues found. Two low-severity observations that are acceptable:

• LOW sh/e2e/lib/clouds/digitalocean.sh:367 — Theoretical integer overflow in wc -l output, mitigated by API pagination (200/page max) and practical droplet limits
• LOW sh/e2e/lib/clouds/digitalocean.sh:368 — Arithmetic could produce negative values if API returns inconsistent data, but handled by minimum-of-1 guard on lines 369-370

Security Assessment

✓ Command injection: Clean - no user input, all data from trusted DigitalOcean API
✓ Credential leaks: Clean - uses _do_curl_auth wrapper with temp file (chmod 600), token never in ps output
✓ Injection safety: Clean - fixed regex patterns, quoted arithmetic expansions
✓ Error handling: Robust - all API calls have fallback to safe default (3)
✓ API response validation: Adequate - uses grep with fixed patterns, handles missing/malformed responses
✓ Integer overflow/underflow: Low risk - bounded by API pagination and practical limits

Tests

• bash -n: PASS
• bun test: PASS (1396 tests, 0 failures)
• curl|bash: N/A (library function)
• macOS compat: OK (uses printf, $(( )), no bash 4+ features)

Code Quality

The implementation is well-structured with appropriate fallbacks and error handling. The dynamic capacity calculation is a significant improvement over the hardcoded limit, reducing E2E test failures when pre-existing droplets consume quota slots.

---

-- security/pr-reviewer

[la14-1]

Summary

• _digitalocean_max_parallel() always returned 3, but pre-existing droplets consume quota slots and cause "droplet limit exceeded" errors when tests run 3-in-parallel
• Now queries /v2/account for the actual droplet_limit and subtracts current droplet count to compute available capacity
• Falls back to 3 if the API is unreachable

Root Cause

The QA account has a 3-droplet limit. During this E2E run, a spawn-5v48 droplet (created 2026-03-11) occupied 1 of 3 slots. Batch 1 tried to create 3 droplets (claude, openclaw, zeroclaw) in parallel — the 3rd slot was already taken, causing claude to fail. Same pattern caused opencode to fail in batch 2.

Test plan

• Verify _digitalocean_max_parallel returns 2 when 1 droplet exists on a 3-limit account
• Verify returns 3 when no pre-existing droplets exist

-- qa/e2e-tester

[louisgv]

Security Review

Verdict: APPROVED
Commit: 1dd6c0c

Findings

No critical or high-severity issues found. Two low-severity observations that are acceptable:

• LOW sh/e2e/lib/clouds/digitalocean.sh:367 — Theoretical integer overflow in wc -l output, mitigated by API pagination (200/page max) and practical droplet limits
• LOW sh/e2e/lib/clouds/digitalocean.sh:368 — Arithmetic could produce negative values if API returns inconsistent data, but handled by minimum-of-1 guard on lines 369-370

Security Assessment

✓ Command injection: Clean - no user input, all data from trusted DigitalOcean API
✓ Credential leaks: Clean - uses _do_curl_auth wrapper with temp file (chmod 600), token never in ps output
✓ Injection safety: Clean - fixed regex patterns, quoted arithmetic expansions
✓ Error handling: Robust - all API calls have fallback to safe default (3)
✓ API response validation: Adequate - uses grep with fixed patterns, handles missing/malformed responses
✓ Integer overflow/underflow: Low risk - bounded by API pagination and practical limits

Tests

• bash -n: PASS
• bun test: PASS (1396 tests, 0 failures)
• curl|bash: N/A (library function)
• macOS compat: OK (uses printf, $(( )), no bash 4+ features)

Code Quality

The implementation is well-structured with appropriate fallbacks and error handling. The dynamic capacity calculation is a significant improvement over the hardcoded limit, reducing E2E test failures when pre-existing droplets consume quota slots.

---

-- security/pr-reviewer

[la14-1]

Summary

• _digitalocean_max_parallel() always returned 3, but pre-existing droplets consume quota slots and cause "droplet limit exceeded" errors when tests run 3-in-parallel
• Now queries /v2/account for the actual droplet_limit and subtracts current droplet count to compute available capacity
• Falls back to 3 if the API is unreachable

Root Cause

The QA account has a 3-droplet limit. During this E2E run, a spawn-5v48 droplet (created 2026-03-11) occupied 1 of 3 slots. Batch 1 tried to create 3 droplets (claude, openclaw, zeroclaw) in parallel — the 3rd slot was already taken, causing claude to fail. Same pattern caused opencode to fail in batch 2.

Test plan

• Verify _digitalocean_max_parallel returns 2 when 1 droplet exists on a 3-limit account
• Verify returns 3 when no pre-existing droplets exist

-- qa/e2e-tester

[louisgv]

Security Review

Verdict: APPROVED
Commit: 1dd6c0c

Findings

No critical or high-severity issues found. Two low-severity observations that are acceptable:

• LOW sh/e2e/lib/clouds/digitalocean.sh:367 — Theoretical integer overflow in wc -l output, mitigated by API pagination (200/page max) and practical droplet limits
• LOW sh/e2e/lib/clouds/digitalocean.sh:368 — Arithmetic could produce negative values if API returns inconsistent data, but handled by minimum-of-1 guard on lines 369-370

Security Assessment

✓ Command injection: Clean - no user input, all data from trusted DigitalOcean API
✓ Credential leaks: Clean - uses _do_curl_auth wrapper with temp file (chmod 600), token never in ps output
✓ Injection safety: Clean - fixed regex patterns, quoted arithmetic expansions
✓ Error handling: Robust - all API calls have fallback to safe default (3)
✓ API response validation: Adequate - uses grep with fixed patterns, handles missing/malformed responses
✓ Integer overflow/underflow: Low risk - bounded by API pagination and practical limits

Tests

• bash -n: PASS
• bun test: PASS (1396 tests, 0 failures)
• curl|bash: N/A (library function)
• macOS compat: OK (uses printf, $(( )), no bash 4+ features)

Code Quality

The implementation is well-structured with appropriate fallbacks and error handling. The dynamic capacity calculation is a significant improvement over the hardcoded limit, reducing E2E test failures when pre-existing droplets consume quota slots.

---

-- security/pr-reviewer