Overview
Theevaluate command performs comprehensive static analysis on a skill directory, running all configured checks and generating a detailed quality report with scores and actionable feedback.
Usage
Arguments
Path to the skill directory. Defaults to current directory if not specified.
Options
Alias:
-oOutput file path for JSON report. Only used when --format json is specified.Alias:
-fOutput format for the report.Options:console- Human-readable terminal output with colors and formattingjson- Machine-readable JSON output
Alias:
-VShow all checks in the report, not just failures. By default, only failed checks are displayed.Alias:
-sOnly run checks required by the Agent Skills specification. Skips quality suggestion checks.Examples
Evaluate current directory
Evaluate specific skill with verbose output
Generate JSON report
Run only spec-required checks
Output Format
Console Output
The default console output includes:- Overall quality score (0-100)
- Pass/fail status
- Results grouped by dimension (structure, naming, content)
- Severity indicators (error, warning, info)
- Actionable error messages
JSON Output
JSON format includes:overall_pass: booleanscore: 0-100checks: array of check results with:check_id: unique identifiercheck_name: human-readable namedimension: structure, naming, description, or contentseverity: error, warning, or infopassed: booleanmessage: description of result
Exit Codes
- 0: Evaluation passed (all checks passed)
- 1: Evaluation failed (one or more checks failed) or error occurred
Notes
The skill directory must contain a
SKILL.md file. If not found, the command will exit with an error.By default,
evaluate runs all 28 checks (10 spec-required + 18 quality suggestions). Use --spec-only to run only the 10 required checks.Related Commands
- sklab validate - Quick pass/fail validation
- sklab list-checks - View all available checks