buildkite-cli

name: buildkite-cli description: "Use the bkci CLI to query Buildkite builds, logs, artifacts, and token scope status with LLM-friendly JSON output."

Buildkite CLI (bkci)

Use this skill when you want CI data from Buildkite through the bkci utility.

This is a good default when you need structured JSON output that is easy for agents to parse.

Install from local checkout

For now, install from a local clone and link it:

git clone https://github.com/PSPDFKit-labs/buildkite-cli <buildkite-cli-dir>
cd <buildkite-cli-dir>
pnpm install
pnpm run build
npm link

Verify:

bkci --help

After pulling updates, rebuild:

pnpm run build

Authentication

Set one of these env vars before calling bkci:

• BUILDKITE_TOKEN
• BUILDKITE_API_TOKEN
• BK_TOKEN

If no env token is set, auth can be configured interactively:

bkci auth setup

This writes ~/.config/buildkite-cli/auth.json with strict permissions.

Agent safety rule

• Do not run bkci auth setup automatically.
• If bkci auth status reports missing token/auth, stop and ask the user to run auth setup manually.
• Do not pass token values in command arguments unless the user explicitly requests it.

Required scopes:

• read_builds
• read_build_logs
• read_artifacts

Validate token/scopes first:

bkci auth status

Common commands

List builds:

bkci builds list --org ORG --pipeline PIPELINE --per-page 10

Get one build with jobs:

bkci builds get --org ORG --pipeline PIPELINE --build BUILD_NUMBER

Fetch one job log (cleaned output):

bkci jobs log get --org ORG --pipeline PIPELINE --build BUILD_NUMBER --job JOB_ID --tail-lines 400 --max-bytes 250000

List artifacts:

bkci artifacts list --org ORG --pipeline PIPELINE --build BUILD_NUMBER

Download artifact(s):

bkci artifacts download --org ORG --pipeline PIPELINE --build BUILD_NUMBER --artifact-id ARTIFACT_ID --out /tmp/bk-artifacts

List annotations:

bkci annotations list --org ORG --pipeline PIPELINE --build BUILD_NUMBER

Output contract

bkci always returns a stable top-level JSON envelope:

• ok
• apiVersion
• command
• request
• summary
• pagination
• data
• error

Use --raw to keep exact Buildkite payloads in data.

Recommended usage pattern

1. auth status (if this fails due to missing token/scopes, ask user to fix auth before continuing)
2. builds list (optionally filtered by --pipeline, --branch, --state)
3. builds get for the selected build
4. jobs log get for relevant job IDs
5. artifacts list / artifacts download
6. annotations list