reach doctor¶
Inspect and diagnose local development environment, runtime agent binaries, API credentials, and skill catalogs.
Synopsis¶
Diagnostics Checked¶
- Environment: Verifies Python
>= 3.12runtime/platform details and checks whether optionalSemantic Scoring (model2vec)(skill-reach[semantic]) is installed. - Runtimes: Checks for installed agent CLI executables (
claude,agy/antigravity,goose,pi), theAntigravity SDK(google.antigravity), and the built-inKeyword Runtime (BM25). - Credentials: Verifies configured API keys (
GEMINI_API_KEYorGOOGLE_API_KEY), Google Cloud Application Default Credentials (GOOGLE_APPLICATION_CREDENTIALSor standardgcloudADC paths plus active Vertex flagsAGY_ADC_AUTH,GOOGLE_GENAI_USE_VERTEXAI,GOOGLE_GENAI_USE_ENTERPRISE), and Google Cloud Agent Registry project and cache state. - Skills: Discovers and counts resident skills across workspace paths (respecting
reach.toml[discovery].precedence,[study].skills, rootSKILL.md, and canonical symlink deduplication), plus optional user-global directories (~/) when--global(-g) is enabled. - Configuration: Validates the TOML syntax, Pydantic schema, and active settings in
reach.toml.
reach doctor exits 0 when all checks pass or warn (OK / WARN) and exits 1 when any check fails (FAIL), making it suitable as a preflight health gate in CI/CD pipelines.
Key Scenarios¶
Include user-level skill directories (~/.agents/skills, ~/.claude/skills, etc.) alongside workspace skills:
Options¶
| Option | Type | Default | Description |
|---|---|---|---|
[PATH], --path, -p |
Path | Current directory | Target workspace directory to inspect. |
--verbose, -v |
Flag | false |
Display detailed diagnostics and recommended remediation steps. |
--quiet, -q |
Flag | false |
Mute the terminal table view. |
--global, -g |
Flag | false |
Discover and inspect skills from user global configuration (~/). |
--format |
Choice | None | Output format: json, jsonl, csv (default: terminal table). |
--config, -c |
Path | reach.toml |
Path to reach.toml configuration file. |