What the agent does
The agent reads what the repository already runs before adding a scanner, so the new workflow fits beside the old ones. In a repository like spf13/cobra-cli, go.mod declares go 1.15 with github.com/spf13/cobra v1.6.1 and github.com/spf13/viper v1.12.0, main.go calls cmd.Execute(), and cmd/ holds root.go, add.go and init.go plus golden-file tests that compare generated cmd/*.go output against cmd/testdata/*.golden. .github/workflows/test.yml has three jobs: golangci-lint on Go 1.17 through golangci/golangci-lint-action@v2, test-unix on Ubuntu and macOS across Go 1.14 to 1.17 running make test build with richgo and gox, and test-win under MSYS2. .github/dependabot.yml already updates both gomod and github-actions weekly, and the Makefile builds with gox for linux and darwin on amd64. There is no CodeQL, no security-events permission and no scheduled scan anywhere.
From that the agent writes .github/workflows/codeql.yml. It triggers on push to main, on pull_request, and on a weekly schedule, with a permissions block of security-events: write, contents: read and actions: read and nothing else. One job on ubuntu-latest checks out with actions/checkout@v7, sets up a toolchain with actions/setup-go@v7 and go-version: stable, then runs github/codeql-action/init@v4 with languages: go and build-mode: autobuild, and github/codeql-action/analyze@v4 with category: "/language:go". autobuild is the mode CodeQL's own README recommends for compiled languages with ordinary build steps, and a go build ./... is as ordinary as it gets; the category keeps these results separate if a second analysis is ever added. The go toolchain step is deliberate: go 1.15 in go.mod is a floor, and the extractor needs a current compiler on the runner, which stable provides.
Pinning is by major tag rather than by commit SHA, and the pull request says why: dependabot.yml already watches github-actions with an open pull request limit of 99, so v4 and v7 are updated the same way actions/checkout@v3.1.0 in test.yml is today, and a SHA pin would need the same Dependabot entry to stay fresh. The agent does not touch test.yml, the .golangci.yml file with its older linter list, or the Makefile; CodeQL is a separate check and should fail or pass on its own.
Validation happens in the sandbox. The agent installs actionlint and runs it on the new file to catch a wrong key or an unknown event before the workflow ever runs on GitHub, then runs go vet ./... and go test ./... so the pull request shows the repository still builds with the toolchain the new job will use. CodeQL itself needs GitHub's runners and the code scanning upload, so the first real result appears on the pull request's own checks once the workflow runs there; the description tells the reviewer to look for the CodeQL check and for the Security tab after merge, and to make sure the repository does not also have code scanning's default setup enabled, since GitHub refuses an advanced-setup upload when it is.
Start it with one API call
POST https://api.phoenix.so/o/{org}/sessions with this body:
{
"agent": "claude",
"input": "Add CodeQL scanning to this Cobra CLI repository. Read go.mod, main.go, cmd/root.go, cmd/add.go, cmd/init.go, cmd/golden_test.go, Makefile, .golangci.yml, .github/workflows/test.yml and .github/dependabot.yml first. Write .github/workflows/codeql.yml that runs on push to main, on pull_request and on a weekly schedule, with permissions limited to security-events write, contents read and actions read, one job on ubuntu-latest that uses actions/checkout@v7, actions/setup-go@v7 with go-version stable, github/codeql-action/init@v4 with languages go and build-mode autobuild, and github/codeql-action/analyze@v4 with category /language:go. Pin by major tag and explain in the pull request that dependabot.yml already tracks github-actions. Do not change test.yml, .golangci.yml or the Makefile. Install actionlint and run it on the new workflow, run go vet ./... and go test ./..., and open a pull request that pastes the actionlint and test output and tells the reviewer where the first CodeQL result will appear and to check that default setup for code scanning is not enabled on the repository.",
"resources": [
{
"url": "https://github.com/your-org/your-app"
}
]
}- Get an API key. In the console, open Settings, then API keys, and create a key for your org. Send it in the Authorization header as a Bearer token.
- Send the request. POST the body in request.json to https://api.phoenix.so/o/{org}/sessions, with your repository as the resources URL. The run starts in its own cloud sandbox.
- Read what came back. Follow the run on its live stream or wait for the run.completed webhook, then read the result and its trajectory.
What comes back
What comes back is a pull request that adds one file, .github/workflows/codeql.yml. The description lists the triggers, the three permissions, the four actions with their major tags, the reason autobuild and a stable Go toolchain were chosen over the go 1.15 floor in go.mod, and the actionlint, go vet and go test output.
The run's record holds the actionlint install and run, the vet and the test run with the golden-file cases. Nothing else in the repository changes, and the existing Test workflow keeps its three jobs as they were.
Before merging, a reviewer checks the repository's code security settings on GitHub: if default setup is on, it is turned off before this workflow runs, or the workflow's upload is refused. Once merged, the weekly schedule and every pull request produce results under the Security tab, and Dependabot's existing github-actions entry keeps the four action tags current.
Run it on every new Go repository
An organisation with many Go CLIs starts one run per repository from its own server: one POST /o/{org}/sessions each, with that repository in resources and a request_id that is a UUID v5 of the repository's full name, so a retried call returns the run already started rather than a second pull request. Each run reads that repository's own workflows and dependabot.yml, so a repository that pins by SHA gets SHA pins and a repository without Dependabot gets a github-actions entry added beside the workflow.
Afterwards, an automation on the Project with a github trigger on workflow_run.failure and the prompt "if the failed workflow is CodeQL, read its log, say whether autobuild failed or the upload was refused, and propose the fix as a comment" turns a red scan into a diagnosis rather than a notification. The trigger adds the event after the prompt; only work by an owner, member or collaborator fires it, never a fork's.
The use case this job follows, with its automation and trigger: Hand off a long task.
Why run it on Rush?
A CodeQL workflow is twenty lines, and the mistakes are in the details nobody reads twice: a missing security-events permission, a build mode that cannot compile the module, a toolchain older than the extractor needs, pins that nothing keeps current, and a default-setup conflict that refuses the upload after merge. The agent reads the repository's existing workflows and Dependabot config and writes the file to match them, then lints it before anyone waits on a runner.
In its own sandbox the agent runs actionlint, go vet and the test suite, and the run keeps that output for the pull request. For an organisation with many Go repositories, each gets its own run and its own pull request, and the pattern is consistent without a template being pasted blindly into each.
Frequently asked questions
Why pin by major tag and not by commit SHA?
Because dependabot.yml already tracks github-actions weekly, so v4 and v7 move the same way the existing actions/checkout@v3.1.0 does. A SHA pin is stricter but goes stale without the same Dependabot entry, and the pull request explains the trade so a team with a SHA policy can switch the four lines.
Why does the workflow set up Go when autobuild compiles the code?
The CodeQL Go extractor runs the build with whatever toolchain the runner has, and go 1.15 in go.mod is a minimum, not a request for that version. actions/setup-go@v7 with go-version: stable gives a current compiler that builds the module and that the extractor supports.
Can the agent run CodeQL itself before opening the PR?
Not usefully. The analysis and the upload to code scanning happen on GitHub's side, so the agent validates the workflow with actionlint and proves the module builds and tests with the same toolchain. The first scan result appears as the CodeQL check on the pull request itself, since the workflow triggers on pull_request.
Will this conflict with the existing lint and test workflow?
No. test.yml is untouched and CodeQL is a separate workflow with its own job and permissions. If the repository already has code scanning's default setup turned on, GitHub refuses uploads from an advanced-setup workflow, which is why the description asks the reviewer to check that setting first.
More Cobra jobs
The same job on other stacks
- Add CodeQL Scanning to a Repository with GitHub Actions (GitHub Actions)
- Add CodeQL Scanning to a Go CLI (Go CLI)
- Add CodeQL Scanning to a Gradle Project (Gradle)