CLI reference
Global options
| Flag | Description |
|---|---|
--missions-dir <dir> |
Directory containing mission definitions (default: built-in) |
-V, --version |
Print version |
-h, --help |
Show help |
Commands
scan
Run the AI agent to scan a repo and produce a plan.
quartermaster scan [options]
| Flag | Default | Description |
|---|---|---|
--mission <name> |
deps |
Mission to run |
--repo-dir <dir> |
. |
Repository directory |
-o, --output <file> |
plan.json |
Output plan file |
--model <model> |
anthropic/claude-sonnet-4-20250514 |
LLM model |
--reasoning-effort <level> |
low, medium, or high |
|
--platform <platform> |
auto | gitlab or github |
--project-url <url> |
auto | Project URL |
--default-branch <branch> |
auto | Default branch |
-v, --verbose |
Verbose logging |
validate
Validate a plan JSON file against the mission's rules.
quartermaster validate [options] <plan-file>
| Flag | Default | Description |
|---|---|---|
--mission <name> |
deps |
Mission (determines which allowlist to check against) |
-v, --verbose |
Verbose logging |
Checks: branch patterns, command allowlist, confidence range, working_dir safety, duplicate detection.
execute
Execute a validated plan.
quartermaster execute [options] <plan-file>
| Flag | Default | Description |
|---|---|---|
--mission <name> |
deps |
Mission (determines which allowlist to use) |
--repo-dir <dir> |
. |
Repository directory |
--execute |
Actually run (without this flag, dry-run) | |
--confidence-threshold <n> |
0.5 |
Skip actions below this confidence |
--platform <platform> |
auto | gitlab or github |
--project-url <url> |
auto | Project URL |
--default-branch <branch> |
auto | Default branch |
-v, --verbose |
Verbose logging |
run
Full pipeline: scan + validate + execute.
quartermaster run [options]
| Flag | Default | Description |
|---|---|---|
--mission <name> |
deps |
Mission to run |
--repo-dir <dir> |
. |
Repository directory |
--model <model> |
anthropic/claude-sonnet-4-20250514 |
LLM model |
--reasoning-effort <level> |
low, medium, or high |
|
--execute |
Actually run (without this flag, dry-run) | |
--confidence-threshold <n> |
0.5 |
Skip actions below this confidence |
--platform <platform> |
auto | gitlab or github |
--project-url <url> |
auto | Project URL |
--default-branch <branch> |
auto | Default branch |
-v, --verbose |
Verbose logging |
Missions
Built-in missions:
| Name | Description |
|---|---|
deps |
Scan for outdated dependencies and create update PRs/issues |
docs-drift |
Detect documentation that has drifted from source code changes |
Select a mission with --mission <name>. Default is deps.
To use custom missions, point --missions-dir at a directory containing mission subdirectories.
Models
Use --model provider/model-id or just the model name (provider auto-detected):
--model anthropic/claude-sonnet-4-20250514 # Anthropic
--model google/gemini-2.5-flash # Google
--model openai/gpt-4o # OpenAI
--model bedrock/converse/arn:aws:... # AWS Bedrock
--model gemini-2.5-flash # auto-detects google
--model claude-sonnet-4-20250514 # auto-detects anthropic
Environment variables
| Variable | Provider |
|---|---|
ANTHROPIC_API_KEY |
Anthropic (Claude) |
GEMINI_API_KEY |
Google (Gemini) |
OPENAI_API_KEY |
OpenAI (GPT) |
LLM_API_KEY |
Any (fallback) |
QUARTERMASTER_GITLAB_TOKEN |
GitLab API |
QUARTERMASTER_GITHUB_TOKEN |
GitHub API |
Platform detection
Priority: CLI flags > CI environment variables > git remote > defaults.
Detected automatically from:
GITLAB_CI/CI_PROJECT_URL(GitLab CI)GITHUB_ACTIONS/GITHUB_REPOSITORY(GitHub Actions)git remote get-url origin(local)