Stats
Actions
Available In
Tags
Plugin
scenario-testing
Probabilistic scenario validation for agentic software. Replaces rigid tests with LLM-judged user-story scenarios, satisfaction scoring across observed trajectories, and a Digital Twin Universe for behavioral clones of third-party services. Designed for codebases where agents and LLMs are first-class design primitives.
What's Inside
Slash Commands10
st:catalog
/catalog
List, search, and manage the scenario catalog — view scenarios by domain, check twin coverage, and show catalog health
st:chaos
/chaos
Configure chaos injection for a digital twin — set failure probabilities, choose profiles, and customize failure modes
st:init
/init
Initialize the scenario catalog, config, and directory structure for a new project
st:report
/report
Display the satisfaction report — per-scenario scores, threshold status, trend analysis, and comparison to previous runs
st:review
/review
Review and refine an existing scenario — check quality, update criteria, bump version
Agents6
satisfaction-judge
/satisfaction-judge
Evaluates individual trajectories against scenario satisfaction criteria and anti-patterns. Produces binary satisfactory/unsatisfactory judgments with reasoning. Uses LLM-as-judge methodology.
scenario-author
/scenario-author
Translates natural-language user stories into structured YAML scenarios with personas, satisfaction criteria, and anti-patterns. References existing scenarios to avoid duplication.
scenario-reviewer
/scenario-reviewer
Reviews authored scenarios for completeness, ambiguity, and testability. Ensures satisfaction criteria are specific enough for LLM judgment. Flags vague or overly rigid scenarios.
trajectory-runner
/trajectory-runner
Executes scenarios against the codebase with digital twins substituted for real services. Records every action, state transition, and outcome as a trajectory. Manages parallel execution for volume runs.
twin-builder
/twin-builder
Analyzes the codebase to discover third-party API usage and generates behavioral clones (digital twins) with API surfaces, state machines, and chaos injection configurations.
Skills3
digital-twin-universe
/digital-twin-universe
Framework for building behavioral clones (digital twins) of third-party services. Covers API surface replication, state machines, chaos injection, and twin composition for validating at volumes and rates impossible against live services.
satisfaction-metrics
/satisfaction-metrics
Framework for measuring, aggregating, and trending satisfaction scores across scenarios. Covers LLM-as-judge methodology, trajectory evaluation, threshold configuration, comparison analysis, and reporting.
scenario-methodology
/scenario-methodology
Framework for authoring, structuring, and managing scenarios — end-to-end user stories validated probabilistically by LLM-as-judge. Covers the holdout principle, scenario anatomy, versioning, composition, and anti-reward-hacking patterns.
Stats
Version0.1.0
LanguageShell
Stars1
Forks1
MaintenanceGood
LicenseMIT
Last Commit
Added
Available In
Safety Signals
Caution
Uses power tools
Uses Bash, Write, or Edit tools
