[Fleet Monitor] petry-projects/.github-private — daily-pr-review-health.yml

🟠 DEGRADED: `petry-projects/.github-private` / `daily-pr-review-health.yml`

Metric	Value
Failure Rate	50% (1 / 2 runs)
Status	DEGRADED
p50 Duration	100s
p95 Duration	100s
Successful Runs	1
Cancelled Runs	0

Workflow: daily-pr-review-health.yml

Last updated by Fleet Monitor run on 2026-06-09.
Threshold: failure rate > 10% over the monitored window.

Dev-Lead Implementation Plan

Issue: #511 — [Fleet Monitor] petry-projects/.github-private — daily-pr-review-health.yml

Root Cause

The daily-pr-review-health.yml health check fails when the Claude CLI hits its daily usage limit. From the failing run (2026-06-09T08:58:14Z):

##[error]Claude invocation failed. CLI output follows:
You've hit your limit · resets 4pm (UTC)

The script calls exit 1 unconditionally on any Claude failure, including transient rate limits that are not indicative of a real PR review agent health problem.

Scope

Add graceful rate-limit handling to scripts/pr_review_health.sh so that a transient Claude usage-cap does not mark the health-check workflow as failed. Emit a ::warning:: annotation instead and exit 0.

Implementation Checklist

scripts/pr_review_health.sh — detect rate limit in Claude error handler block; emit ::warning::, override HAS_FAILURES=false, exit 0
tests/test_pr_review_health.bats — new test file covering rate-limit exit-0, non-rate-limit exit-1, and no-failures skip-Claude paths
.github/workflows/lint.yml — add new test file to bats test list (per AGENTS.md exception rule)

Tests to Write

"Claude rate limit exits 0 and emits ::warning:: annotation" — stub claude to output "You've hit your limit", verify exit 0 and ::warning:: in output
"Non-rate-limit Claude failure exits 1 and emits ::error:: annotation" — stub claude to output generic error, verify exit 1 and ::error:: in output
"No failed runs exits 0 without invoking Claude" — stub gh to return all-success runs, verify exit 0 and no Claude invocation

Test Command

bats tests/test_pr_review_health.bats

Pattern Reference

The rate-limit grep pattern follows _rate_limit_pattern() in scripts/engine.sh (the codebase's single source of truth), specifically the hit your limit and resets [0-9]+(am|pm) tokens that match the observed error message.

Dev-Lead: Implementation Complete

Root Cause

The failing run (2026-06-09T08:58:14Z) hit the Claude daily usage limit:

You've hit your limit · resets 4pm (UTC)

scripts/pr_review_health.sh called exit 1 unconditionally on any Claude failure, so a transient rate-limit caused the health-check workflow to register as a hard failure — inflating the failure rate with noise unrelated to the PR review agent's health.

Plan Execution

scripts/pr_review_health.sh — added rate-limit detection after Claude failure; emits ::warning::, overrides HAS_FAILURES=false in $GITHUB_ENV, exits 0
tests/test_pr_review_health.bats — 5 new tests (3 rate-limit paths, 1 non-rate-limit error, 1 no-failures skip)
.github/workflows/lint.yml — added tests/test_pr_review_health.bats to bats test list per AGENTS.md exception rule

Test Results

1..179
...
ok 175 Claude rate limit exits 0 and emits ::warning:: annotation
ok 176 Claude rate limit: HAS_FAILURES is overridden to false in GITHUB_ENV
ok 177 Claude 'usage limit' message also triggers rate-limit handling (exit 0)
ok 178 non-rate-limit Claude failure exits 1 and emits ::error:: annotation
ok 179 no failed runs in window exits 0 without invoking Claude

179 tests, 0 failures

Lint: shellcheck --severity=warning on all 40 scripts/**/*.sh — passed. Agent profile frontmatter — passed.

Files Changed

scripts/pr_review_health.sh: 13-line addition in the Claude error handler — detect rate-limit keywords via grep -qiE (pattern mirrors _rate_limit_pattern() in engine.sh), emit ::warning::, override HAS_FAILURES=false, exit 0
tests/test_pr_review_health.bats: new test file, 5 tests covering all error-handling paths
.github/workflows/lint.yml: added new test file to bats run list

Notes

The HAS_FAILURES=false override prevents the issue-creation step from filing a spurious ticket; a ::warning:: annotation in the Actions log preserves visibility for operators
Non-rate-limit Claude errors (e.g. segfault, configuration error) still exit 1 as before — only genuine rate-limit/quota messages are handled gracefully
Pattern breadth: quota alone is intentionally conservative; combined with the Claude-specific context (only fires after if ! claude ...) false-positive risk is negligible

feat: implement issue #511 — [Fleet Monitor] petry-projects/.github-p…

Navigation Menu

Repository navigation

[Fleet Monitor] petry-projects/.github-private — daily-pr-review-health.yml #511

🟠 DEGRADED: `petry-projects/.github-private` / `daily-pr-review-health.yml`

Dev-Lead Implementation Plan

Root Cause

Scope

Implementation Checklist

Tests to Write

Test Command

Pattern Reference

Dev-Lead: Implementation Complete

Root Cause

Plan Execution

Test Results

Files Changed

Notes

15 remaining items

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

[Fleet Monitor] petry-projects/.github-private — daily-pr-review-health.yml #511

Description

🟠 DEGRADED: petry-projects/.github-private / daily-pr-review-health.yml

Activity

don-petry commented on Jun 9, 2026

Dev-Lead Implementation Plan

Root Cause

Scope

Implementation Checklist

Tests to Write

Test Command

Pattern Reference

don-petry commented on Jun 9, 2026

Dev-Lead: Implementation Complete

Root Cause

Plan Execution

Test Results

Files Changed

Notes

15 remaining items

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

🟠 DEGRADED: `petry-projects/.github-private` / `daily-pr-review-health.yml`