Skip to main content

​
Overview

The trigger command runs trigger tests to verify that a skill activates correctly for different types of prompts. It tests whether the skill is properly discovered and invoked by the agent runtime.

​
Usage

sklab trigger [SKILL_PATH] [OPTIONS]

​
Arguments

​
SKILL_PATH
Path
Path to the skill directory. Defaults to current directory if not specified.

​
Options

​
--type
string
Alias: -tOnly run tests of a specific trigger type. Filters test execution.Options:
  • explicit - Skill named directly with $ prefix
  • implicit - Describes exact scenario without naming skill
  • contextual - Realistic noisy prompt with domain context
  • negative - Should NOT trigger (catches false positives)
sklab trigger --type explicit
​
--output
Path
Alias: -oOutput file path for JSON report.
sklab trigger --output trigger-report.json
​
--format
string
default:"console"
Alias: -fOutput format for the report.Options:
  • console - Human-readable terminal output with table formatting
  • json - Machine-readable JSON output
sklab trigger --format json

​
Trigger Types

The command tests four types of skill activation:

​
Explicit

Skill named directly with $ prefix in the prompt: 
Use $my-skill to analyze this data

​
Implicit

Describes the exact scenario without naming the skill: 
Analyze this customer feedback data for sentiment patterns

​
Contextual

Realistic prompt with domain context and noise: 
Our support team received 500 tickets this week. Can you help me understand
the main themes and customer sentiment trends?

​
Negative

Should NOT trigger the skill (prevents false positives): 
Show me the sales report from last quarter

​
Examples

​
Run all trigger tests

sklab trigger

​
Run only explicit trigger tests

sklab trigger --type explicit

​
Run tests with JSON output

sklab trigger --format json --output results.json

​
Test specific skill

sklab trigger ./skills/sentiment-analysis

​
Output Format

​
Console Output

Displays a formatted table with:
  • Test name
  • Trigger type
  • Pass/fail status (✓ or ✗)
  • Summary statistics by type
  • Overall pass rate and duration
Trigger Test Report: sentiment-analysis
Runtime: claude │ Duration: 12.3s │ 8/10 passed

┌─────────────────────────┬────────────┬────────┐
│ Test                    │ Type       │ Status │
├─────────────────────────┼────────────┼────────┤
│ direct-invocation       │ explicit   │   ✓    │
│ scenario-description    │ implicit   │   ✓    │
│ real-world-request      │ contextual │   ✓    │
│ unrelated-request       │ negative   │   ✗    │
└─────────────────────────┴────────────┴────────┘

By type: explicit: 3/3 (100%) │ implicit: 2/3 (67%) │ contextual: 2/3 (67%) │ negative: 1/1 (100%)

​
JSON Output

Includes:
  • skill_name: name of the skill tested
  • runtime: runtime used (e.g., “claude”)
  • tests_run: total number of tests
  • tests_passed: number of passed tests
  • overall_pass: boolean
  • duration_ms: execution time in milliseconds
  • results: array of test results
  • summary_by_type: statistics grouped by trigger type

​
Test Definition Files

Trigger tests must be defined in YAML files located at:
  • .skill-lab/tests/scenarios.yaml (legacy)
  • .skill-lab/tests/triggers.yaml (recommended)
If no test files are found, the command suggests using sklab generate to create them automatically.

​
Exit Codes

  • 0: All trigger tests passed
  • 1: One or more tests failed or error occurred

​
Notes

Trigger tests require test definitions in .skill-lab/tests/triggers.yaml. Use sklab generate to auto-generate tests.
The Claude runtime requires the Claude CLI to be properly configured. Ensure you have access to Claude’s API.

Build docs developers (and LLMs) love

Get started for freeTalk to us