Fetch CI Build

Overview

Fetch CI build results, diagnose failures, extract actionable error information, and suggest fixes. Supports multiple CI providers with automatic detection.

Supported Providers

Auto-Detection

From URL

If the user provides a CI URL, detect provider:

• github.com/.../actions/runs/... → GitHub Actions
• buildkite.com/... → Buildkite
• app.circleci.com/... or circleci.com/... → CircleCI

From Project Files

Check for CI configuration directories:

# GitHub Actions
test -d .github/workflows && echo "github"

# Buildkite
test -d .buildkite && echo "buildkite"

# CircleCI
test -d .circleci && echo "circleci"

Workflow

digraph workflow {
    rankdir=TB;
    node [shape=box];
    
    detect [label="1. Detect CI provider"];
    load [label="2. Load provider reference"];
    fetch [label="3. Fetch build results"];
    check [label="4. Check for failures" shape=diamond];
    passed [label="Report: Build passed!"];
    read [label="5. Read failing source files"];
    present [label="6. Present failures + proposed fixes"];
    ask [label="7. Ask: Apply fix?" shape=diamond];
    apply [label="Apply the fix"];
    complex [label="Complex failure?" shape=diamond];
    debug [label="Use systematic-debugging skill"];
    next [label="Next failure?" shape=diamond];
    done [label="Done"];
    
    detect -> load;
    load -> fetch;
    fetch -> check;
    check -> passed [label="passed"];
    check -> read [label="failed"];
    read -> present;
    present -> ask;
    ask -> apply [label="yes"];
    ask -> complex [label="no"];
    apply -> next;
    complex -> debug [label="yes"];
    complex -> next [label="no"];
    debug -> next;
    next -> read [label="yes"];
    next -> done [label="no"];
}

Step-by-Step Process

1. Detect CI Provider

First, check if the user provided a URL. If not, detect from project files.

2. Load Provider Reference

Read the appropriate reference file for provider-specific commands:

• GitHub: references/github.md
• Buildkite: references/buildkite.md
• CircleCI: references/circleci.md

3. Fetch Build Results

Use the provider-specific commands to fetch build information and failures.

4. For Each Failure

Read the relevant source file to understand context:

• For test failures: read the test file at the indicated line
• For lint errors: read the source file at the indicated line
• For TypeScript errors: read the file and understand the type issue

5. Present Findings and Propose Fix

Show the user:

• What failed (test name, file, line)
• Error message
• Proposed fix based on the error

6. Ask User What To Do

Ask how to proceed:

• "Apply the suggested fix"
• "Investigate further before fixing"
• "Skip this failure"
• "Use systematic-debugging for deeper investigation"

7. Complex Failures

If the failure requires deeper investigation (e.g., unclear root cause, flaky test, environmental issue), recommend the systematic-debugging skill.

Error Types Detected

Common Mistakes

Integration with Other Skills

• systematic-debugging: Use for complex failures requiring root cause analysis
• test-driven-development: After fixing, ensure tests follow TDD principles
• verification-before-completion: Run tests locally before pushing fix