Windsurf CI Integration
Overview
Integrate Windsurf configuration validation and AI code quality gates into CI/CD pipelines. Covers validating .devin/rules/project.md, enforcing team policies for AI-generated code, and automating Windsurf config distribution.
Prerequisites
- GitHub repository with Actions enabled
- Windsurf configuration files in repository
- Team agreement on AI code review policy
Tool Use
- Use
Readto inspect only the repository files and configuration needed for the request. - Use
Writeonly for a new artifact the user requested; never write credentials or unreviewed production configuration. - Use
Editfor bounded, reviewable changes and preserve unrelated user work. - Use only the command-scoped
Bashentries declared in frontmatter, with non-destructive checks before mutations.
Instructions
Step 1: Validate Windsurf Config in CI
# .github/workflows/windsurf-config.yml
name: Windsurf Config Validation
on:
pull_request:
paths:
- '.devin/rules/project.md'
- '.codeiumignore'
- '.windsurf/**'
jobs:
validate:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- name: Check .devin/rules/project.md exists and is valid
run: |
if [ ! -f .devin/rules/project.md ]; then
echo "::error::.devin/rules/project.md is missing"
exit 1
fi
CHARS=$(wc -c < .devin/rules/project.md)
if [ "$CHARS" -gt 12000 ]; then
echo "::error::.devin/rules/project.md exceeds 12000 char limit ($CHARS chars)"
exit 1
fi
echo ".devin/rules/project.md: $CHARS chars (limit: 12000)"
- name: Check .codeiumignore covers secrets
run: |
REQUIRED_PATTERNS=(".env" "*.pem" "*.key" "credentials")
MISSING=()
for pattern in "${REQUIRED_PATTERNS[@]}"; do
if ! grep -q "$pattern" .codeiumignore 2>/dev/null; then
MISSING+=("$pattern")
fi
done
if [ ${#MISSING[@]} -gt 0 ]; then
echo "::warning::.codeiumignore missing patterns: ${MISSING[*]}"
fi
- name: Validate workspace rules frontmatter
run: |
for rule in .devin/rules/*.md; do
[ -f "$rule" ] || continue
if ! head -1 "$rule" | grep -q "^---"; then
echo "::error::$rule missing YAML frontmatter"
exit 1
fi
# Check for required trigger field
if ! grep -q "^trigger:" "$rule"; then
echo "::warning::$rule missing 'trigger:' in frontmatter"
fi
done
Step 2: AI Code Quality Gate
# .github/workflows/ai-code-review.yml
name: AI Code Quality Gate
on: pull_request
jobs:
ai-review:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
with: { fetch-depth: 0 }
- name: Detect large AI-generated changesets
run: |
FILES_CHANGED=$(git diff --name-only origin/main..HEAD | wc -l)
if [ "$FILES_CHANGED" -gt 20 ]; then
echo "::warning::Large changeset ($FILES_CHANGED files). If AI-generated, ensure thorough review."
fi
- name: Enforce tests for new source files
run: |
NEW_SRC=$(git diff --name-only --diff-filter=A origin/main..HEAD | grep -cE '\.(ts|js|tsx|jsx)$' || true)
NEW_TEST=$(git diff --name-only --diff-filter=A origin/main..HEAD | grep -cE '\.(test|spec)\.' || true)
if [ "$NEW_SRC" -gt 3 ] && [ "$NEW_TEST" -eq 0 ]; then
echo "::error::$NEW_SRC new source files added without tests"
exit 1
fi
- name: Check for hardcoded secrets in new files
run: |
git diff origin/main..HEAD -- '*.ts' '*.js' '*.tsx' '*.jsx' | \
grep -E '(sk_live|sk_test|AKIA|ghp_|glpat-|xoxb-)' && {
echo "::error::Potential hardcoded secret detected"
exit 1
} || true
Step 3: Distribute Windsurf Config Templates
# .github/workflows/sync-windsurf-config.yml
name: Sync Windsurf Config
on:
push:
branches: [main]
paths: ['windsurf-templates/**']
jobs:
distribute:
runs-on: ubuntu-latest
strategy:
matrix:
repo: [frontend, backend, mobile]
steps:
- uses: actions/checkout@v4
- name: Open reviewed config update in child repos
run: |
echo "Use an organization-owned GitHub App or reusable workflow to create"
echo "a branch and PR in ${{ matrix.repo }}; do not write directly to its default branch."
gh workflow run sync-devin-rules.yml \
--repo "${{ github.repository_owner }}/${{ matrix.repo }}" \
--field source_sha="${{ github.sha }}"
env:
GH_TOKEN: ${{ secrets.REPO_SYNC_TOKEN }}
Step 4: Cascade-Generated Commit Convention
Enforce commit message conventions for AI-generated code:
# In branch protection or CI
- name: Check AI commit convention
run: |
COMMITS=$(git log origin/main..HEAD --pretty=format:"%s")
# If PR has many file changes, warn about AI commit tagging
FILES=$(git diff --stat origin/main..HEAD | tail -1 | awk '{print $1}')
if [ "$FILES" -gt 10 ]; then
if ! echo "$COMMITS" | grep -q "\[cascade\]"; then
echo "::notice::Large changeset without [cascade] tag. If AI-generated, tag commits with [cascade] prefix."
fi
fi
Step 5: MCP Server Health Check (Optional)
- name: Validate MCP config
run: |
MCP_CONFIG="$HOME/.codeium/windsurf/mcp_config.json"
if [ -f "$MCP_CONFIG" ]; then
python3 -c "import json; json.load(open('$MCP_CONFIG'))" || {
echo "::error::MCP config is invalid JSON"
exit 1
}
fi
Output
Produce a reviewable CI change set containing the policy checks, protected-branch-safe workflow, local reproduction commands, expected pass and failure evidence, and rollback instructions. Never push directly to a protected default branch or embed repository credentials in generated files.
Error Handling
| Issue | Cause | Solution |
|---|---|---|
| .devin/rules/project.md over limit | Too many rules | Split into workspace rules in .devin/rules/ |
| Secret detected in diff | AI generated hardcoded key | Remove, rotate, add to .codeiumignore |
| Config sync fails | Token lacks repo access | Update REPO_SYNC_TOKEN permissions |
| Frontmatter validation fails | Missing trigger field | Add trigger: always_on or appropriate mode |
Examples
Branch Protection Rules
# Recommended for teams using Windsurf Cascade
required_status_checks:
- "windsurf-config"
- "ai-code-review"
- "test"
Pre-Commit Hook for .devin/rules/project.md
#!/bin/bash
# .git/hooks/pre-commit
CHARS=$(wc -c < .devin/rules/project.md 2>/dev/null || echo 0)
if [ "$CHARS" -gt 12000 ]; then
echo "ERROR: .devin/rules/project.md exceeds 12000 char limit ($CHARS chars)"
exit 1
fi
Resources
- Focused first-party references
- GitHub Actions Documentation
- Windsurf Admin Guide
Related Skill
Continue with windsurf-deploy-integration to connect these validation gates to a reviewed preview and production deployment workflow.