Reference documentation for Claude Code CI/CD integrations -- GitHub Actions (@claude mentions, workflow setup, action parameters, AWS Bedrock and Google Vertex AI), GitLab CI/CD (pipeline configuration, provider abstraction, OIDC authentication), and Claude Code in Slack (routing modes, session flow, repository selection, channel access control).
This skill provides the complete official documentation for Claude Code integrations with CI/CD platforms and collaboration tools — GitHub Actions, GitLab CI/CD, GitHub Enterprise Server, automated Code Review, and Claude Code in Slack.
Setup options:
/install-github-app in Claude terminal (Claude API only)ANTHROPIC_API_KEY secret, copy workflow from examplesAction parameters (v1):
| Parameter | Description | Required |
| :-------------------- | :----------------------------------------------------------------- | :-------- |
| prompt | Instructions or skill invocation (plain text or /skill-name) | No |
| claude_args | CLI arguments passed through to Claude Code | No |
| plugin_marketplaces | Newline-separated plugin marketplace git URLs | No |
| plugins | Newline-separated plugin names to install before execution | No |
| anthropic_api_key | Claude API key (required for direct API; not used with Bedrock/Vertex) | Conditional |
| github_token | GitHub token for API access | No |
| trigger_phrase | Custom trigger phrase (default: @claude) | No |
| use_bedrock | Use Amazon Bedrock instead of Claude API | No |
| use_vertex | Use Google Vertex AI instead of Claude API | No |
Common claude_args flags:
| Flag | Description |
| :----------------- | :----------------------------------------- |
| --max-turns | Maximum conversation turns (default: 10) |
| --model | Model to use |
| --mcp-config | Path to MCP configuration |
| --allowedTools | Comma-separated allowed tools list |
| --debug | Enable debug output |
Beta → v1 breaking changes:
| Old Beta Input | New v1.0 Input |
| :-------------------- | :------------------------------------ |
| mode | (Removed — auto-detected) |
| direct_prompt | prompt |
| override_prompt | prompt with GitHub variables |
| custom_instructions | claude_args: --append-system-prompt |
| max_turns | claude_args: --max-turns |
| model | claude_args: --model |
| allowed_tools | claude_args: --allowedTools |
| disallowed_tools | claude_args: --disallowedTools |
| claude_env | settings JSON format |
Cloud provider secrets:
| Provider | Required Secrets |
| :---------------- | :----------------------------------------------------------------- |
| Claude API | ANTHROPIC_API_KEY |
| Amazon Bedrock | AWS_ROLE_TO_ASSUME, plus optional APP_ID / APP_PRIVATE_KEY |
| Google Vertex AI | GCP_WORKLOAD_IDENTITY_PROVIDER, GCP_SERVICE_ACCOUNT, plus optional APP_ID / APP_PRIVATE_KEY |
Bedrock model IDs include a region prefix, e.g. us.anthropic.claude-sonnet-4-6.
Setup:
ANTHROPIC_API_KEY as a masked CI/CD variable (Settings → CI/CD → Variables).gitlab-ci.yml using node:24-alpine3.21; install Claude via curl -fsSL https://claude.ai/install.sh | bashKey job variables:
| Variable | Purpose |
| :--------------- | :------------------------------------------------------------ |
| AI_FLOW_INPUT | Prompt / instruction for Claude (set by event listener) |
| AI_FLOW_CONTEXT | Context payload from trigger |
| AI_FLOW_EVENT | Event name that triggered the job |
| AWS_ROLE_TO_ASSUME | IAM role ARN (Bedrock) |
| AWS_REGION | Bedrock region |
| GCP_WORKLOAD_IDENTITY_PROVIDER | WIF provider resource name (Vertex AI) |
| GCP_SERVICE_ACCOUNT | Service account email (Vertex AI) |
| CLOUD_ML_REGION | Vertex AI region (e.g. us-east5) |
Trigger phrase: @claude in issue or MR comments; requires a webhook/listener to call the pipeline trigger API.
Recommended script invocation:
claude -p "${AI_FLOW_INPUT:-'default task'}" \
--permission-mode acceptEdits \
--allowedTools "Bash Read Edit Write mcp__gitlab" \
--debug
Trigger modes (per repo):
| Mode | When reviews run |
| :--------------------- | :-------------------------------------------------------------- |
| Once after PR creation | On PR open / ready-for-review |
| After every push | Every push to the PR branch |
| Manual | Only on explicit @claude review or @claude review once comment |
Manual trigger commands (top-level PR comment only):
| Command | Effect |
| :-------------------- | :------------------------------------------------------------------ |
| @claude review | Runs review; subscribes PR to push-triggered reviews going forward |
| @claude review once | Runs a single review without subscribing to future pushes |
Severity levels:
| Marker | Level | Meaning | | :----- | :---------- | :--------------------------------------------------------- | | Red circle | Important | A bug that should be fixed before merging | | Yellow circle | Nit | Minor issue, worth fixing but not blocking | | Purple circle | Pre-existing | Bug in codebase not introduced by this PR |
Customization files:
| File | Scope | Priority |
| :---------- | :--------------------- | :---------------------- |
| CLAUDE.md | All Claude Code tasks | Violations flagged as nits |
| REVIEW.md | Code Review only | Highest — injected directly into every agent |
REVIEW.md can tune: severity definitions, nit volume cap, skip rules (paths, branches, categories), repo-specific checks, verification bar, re-review convergence, and summary shape.
Parse severity counts from check run output:
gh api repos/OWNER/REPO/check-runs/CHECK_RUN_ID \
--jq '.output.text | split("bughunter-severity: ")[1] | split(" -->")[0] | fromjson'
Returns {"normal": N, "nit": N, "pre_existing": N} — normal holds Important count.
Pricing: ~$15–25 per review; billed via usage credits, separate from plan usage. Set a monthly spend cap at claude.ai/admin-settings/usage.
Local review: use the /code-review command in any Claude Code session. Pass --comment to post inline PR comments, --fix to apply findings. /code-review ultra --fix runs the cloud-based ultrareview.
Feature support:
| Feature | GHES support | Notes |
| :------------------- | :-------------- | :--------------------------------------------------------- |
| Claude Code on web | Supported | Admin connects once; claude --remote auto-detects GHES |
| Code Review | Supported | Same as github.com |
| Claude Security | Supported | Public beta for Enterprise at claude.ai/security |
| Teleport sessions | Supported | --teleport works with GHES repos |
| Plugin marketplaces | Supported | Use full git URLs instead of owner/repo shorthand |
| GitHub Actions | Supported | Manual workflow setup only; /install-github-app is github.com only |
| GitHub MCP server | Not supported | Use gh CLI configured for GHES host instead |
GitHub App permissions required:
| Permission | Access | Used for | | :--------------- | :------------- | :---------------------------------------- | | Contents | Read and write | Cloning repos, pushing branches | | Pull requests | Read and write | Creating PRs, posting review comments | | Issues | Read and write | Responding to issue mentions | | Checks | Read and write | Posting Code Review check runs | | Actions | Read | Reading CI status for auto-fix | | Repository hooks | Read and write | Webhooks for contribution metrics | | Metadata | Read | Required by GitHub for all apps |
GHES plugin marketplace (full git URL required):
/plugin marketplace add git@github.example.com:platform/claude-plugins.git
# or
/plugin marketplace add https://github.example.com/platform/claude-plugins.git
Allowlist GHES in managed settings using hostPattern:
{
"strictKnownMarketplaces": [
{ "source": "hostPattern", "hostPattern": "^github\\.example\\.com$" }
]
}
Network: GHES must be reachable from Anthropic infrastructure; allowlist Anthropic API IP addresses if behind a firewall.
Requirements:
| Requirement | Details |
| :------------------- | :------------------------------------------------------------------------- |
| Claude Plan | Pro, Max, Team, or Enterprise with Claude Code access |
| Claude Code on web | Must be enabled at claude.ai/code |
| GitHub account | Connected to Claude Code on web with at least one authenticated repository |
| Slack authentication | Slack account linked to Claude account via the Claude app |
Routing modes:
| Mode | Behavior | | :---------- | :------------------------------------------------------------------------------ | | Code only | All @mentions go to Claude Code sessions | | Code + Chat | Claude routes each message to Code (coding tasks) or Chat (general questions) |
Session flow: mention @Claude → coding intent detected → Code session created on claude.ai/code → status updates posted in thread → completion summary with "View Session" and "Create PR" buttons.
Limitations:
Channel setup: invite with /invite @Claude — Claude only responds in channels where it has been added.
For the complete official documentation, see the reference files:
Search for places (restaurants, cafes, etc.) via Google Places API proxy on localhost.
Interact with GitHub using the `gh` CLI. Use `gh issue`, `gh pr`, `gh run`, and `gh api` for issues, PRs, CI runs, and advanced queries.
Create or update AgentSkills. Use when designing, structuring, or packaging skills with scripts, references, and assets.
Start voice calls via the OpenClaw voice-call plugin.
Notion API for creating and managing pages, databases, and blocks.
Gemini CLI for one-shot Q&A, summaries, and generation.
Category:developer