skill-tester

name	skill-tester
description	Run local tests for each OpenClaw skill before publishing to ClawHub. Discovers skills, runs configured or heuristic tests, reports pass/fail. Use to ensure all skills work as intended before upload.
version	1.0.0

Skill Tester

Description

Run local tests for each OpenClaw skill before publishing to ClawHub. Discovers skills, runs configured or heuristic tests, reports pass/fail. Use to ensure all skills work as intended before upload.

Skill Tester

Run robust local tests on your OpenClaw skills before uploading to ClawHub. Discovers skills in workspace/skills, runs test commands (from skill_tests.json or heuristics), and outputs a pass/fail report.

Usage

• Before publishing skills to ClawHub: run --all to verify every skill.
• After changing a skill: run --skill SLUG to test that one.
• In CI or as a sub-agent task: use --json for machine-readable report.

python3 <skill-dir>/scripts/skill_tester.py --list
python3 <skill-dir>/scripts/skill_tester.py --all [--json] [-v]
python3 <skill-dir>/scripts/skill_tester.py --skill <slug> [--json] [-v]

• --list — List discoverable skills (have SKILL.md or _meta.json). No tests run.
• --all — Run tests for every discovered skill.
• --skill SLUG — Run tests only for the given skill (e.g. agent-swarm, gateway-guard).
• --json — Output a JSON report: { "skills": { ... }, "passed": N, "failed": M }. Exit 1 if any failed.
• -v / --verbose — Print each test name and result per skill.

Test config

Tests are defined in scripts/skill_tests.json (next to skill_tester.py). Each key is a skill slug; value is a list of test objects:

• name — Test label in output.
• cmd — List of command + args (e.g. ["python3", "scripts/router.py", "spawn", "--json", "hello"]).
• cwd — Optional subdir of the skill (default: skill root).
• expect_exit — Expected process exit code (default 0).
• expect_json_keys — Optional list of keys that must exist in stdout JSON.
• timeout — Seconds (default 25).

If a skill has no entry in skill_tests.json, the tester runs a heuristic: for each scripts/*.py, try --help or --json and treat success as pass.

Requirements

• OPENCLAW_HOME set (or ~/.openclaw) with workspace/skills/ containing skill dirs.
• Python 3.6+.
• Skills under test must have their own dependencies (e.g. gateway-guard needs openclaw CLI for full ensure; status --json can still be tested).

Exit codes

• 0 — All tested skills passed.
• 1 — At least one test failed.
• 2 — Usage error (e.g. missing --skill/--all, or skills dir not found).