analyze-ci

name	analyze-ci
description	Analyze CI failures by fetching logs and producing a structured fix plan. Use before attempting automated CI fixes.

CI Failure Analysis Skill

You are given a list of failed CI checks. Fetch the actual logs, identify the root cause of each failure, and produce a structured fix plan.

Workflow

Step 0 — Check for prior attempts (always do this first)

Before downloading any logs, check whether a previous fix attempt already ran:

1. If .forge/fix-plan.md exists, read it. Identify which failures were previously fixable and which were skipped, and what fix was applied.
2. Run git log --oneline -10 to see what changes were committed by prior fix attempts.
3. For each test still failing: was it fixed before? If yes, the prior fix did not work — do not repeat it.

Step 1 — Fetch logs

Create .forge/logs/ if it doesn't exist: mkdir -p .forge/logs

For each failed check, download the log:

# Generic URL (works for most CI systems)
curl -sL "{log_url}" -o .forge/logs/{check-name}.txt

# GitHub Actions job log
gh api repos/{owner}/{repo}/actions/jobs/{job-id}/logs > .forge/logs/{check-name}.txt

# Compressed archive
curl -sL "{url}" -o .forge/logs/{check-name}.tar.gz
tar -xzf .forge/logs/{check-name}.tar.gz -C .forge/logs/{check-name}/

Analyze downloaded files locally — do not print large log content to the conversation.

Step 2 — Categorize each failure

Fixable by code change:

• compile — build or compilation errors with clear error messages
• lint — lint rule violations (ruff, golangci-lint, eslint, etc.)
• format — formatting violations (gofmt, ruff format, prettier)
• codegen-outdated — generated files out of sync with source
• unit-test — test assertion failures caused by a code bug
• e2e-code-bug — end-to-end test fails consistently with the same assertion error pointing to a logic defect in the code under test

Not fixable by code change — skip:

• infra — CI infrastructure failures (runner unavailable, network timeout, quota exceeded)
• flaky — non-deterministic failures with varying errors across runs

Step 3 — Write the fix plan

Write .forge/fix-plan.md:

# CI Fix Plan

## Summary
[1-2 sentences: what failed and what the fix involves]

## Fixable Failures

### {check-name}
**Category**: {compile | lint | format | codegen-outdated | unit-test | e2e-code-bug}
**Root Cause**: {exact error or description}
**Affected Files**: {list}
**Fix**:
1. {exact command or edit}
2. {verification command}

## Skipped Failures

### {check-name}
**Reason**: {infra | flaky} — {brief explanation}

Important

• Fetch the actual logs — do not guess based on the check name
• Be specific: include exact file paths and error messages
• If a log is unavailable, mark the failure as skipped with reason "log unavailable"
• If a test was previously fixed and is still failing, do not repeat the same approach — re-read the code and find a different root cause

Documentation ripple

For every fixable failure that changes a constant, threshold, or behavior, search for stale references:

grep -r "<old value>" . --include="*.go" --include="*.py" --include="*.md" -l

Include stale documentation files in Affected Files alongside the implementation files.