Claude Code Plugin
Omen provides a Claude Code plugin that packages common analysis workflows into skills -- predefined sequences of tool calls that combine multiple analyzers to answer high-level questions about a codebase.
Installation
/plugin install panbanda/omen
After installation, verify the skills are available:
/skills
You should see the Omen skills listed among your available skills.
Prerequisites
The plugin requires Omen's MCP server to be configured. If you haven't already set it up:
claude mcp add omen -- omen mcp
The plugin invokes Omen tools through the MCP server. Without it, skills will fail with a connection error.
Available Skills
The plugin provides 11 skills that cover the most common analysis workflows.
setup-config
Generate an omen.toml configuration file tailored to the current project. Analyzes the codebase to detect languages, project structure, and test frameworks, then produces a configuration with appropriate defaults.
/setup-config
Creates omen.toml in the project root with analyzer thresholds, excluded paths, and score weights calibrated for the detected project type.
onboard-codebase
Analyze a codebase you are unfamiliar with. Runs repository score, complexity, dependency graph, ownership, and repository map to produce a structured overview of the project's architecture, health, and key areas of concern.
/onboard-codebase
Useful when you are starting work on a new project or reviewing a repository for the first time. The output covers: overall health score, most complex modules, dependency structure, primary contributors per area, and areas of concern.
check-quality
Run quality gates and report pass/fail status. Checks the repository score against the configured threshold and breaks down component scores so you can see which dimensions are dragging the overall score down.
/check-quality
Returns the overall score, per-component breakdown, and whether the configured minimum threshold is met. Equivalent to omen score but with interpreted results and recommended actions.
find-bugs
Identify areas of the codebase most likely to contain defects. Combines defect prediction, hotspot analysis, and complexity data to produce a ranked list of high-risk files and functions.
/find-bugs
The output prioritizes files that score high on multiple risk signals simultaneously: high complexity, frequent changes, low test coverage (via mutation testing), and prior defect history. Files that appear as hotspots across multiple dimensions are the highest priority for review and testing.
focus-review
Prioritize code review effort. Given the current set of changes (uncommitted or a branch diff), identifies which changed files carry the most risk and recommends review order.
/focus-review
Uses diff analysis, churn history, ownership data, and complexity to rank changed files by risk. Files modified by a single contributor with no prior ownership are flagged as higher risk (the "bus factor" concern).
audit-flags
Check feature flag hygiene. Reports all detected feature flags, their staleness status, and recommends flags for cleanup.
/audit-flags
Returns a list of all flags grouped by provider, with stale flags highlighted and sorted by age. Includes the total number of references for each flag, making it easier to estimate cleanup effort.
audit-architecture
Detect architectural issues including dependency cycles, god classes, high coupling, low cohesion, and dead code. Combines dependency graph analysis, CK metrics, code smells, and dead code detection.
/audit-architecture
The output identifies specific architectural problems with locations and recommended actions. Dependency cycles are reported with the full cycle path. God classes are identified by WMC and method count. High-coupling modules are listed with their afferent and efferent coupling counts.
plan-refactoring
Prioritize refactoring work by combining technical debt gradient, complexity, code smells, and churn data. Produces a ranked list of refactoring targets ordered by impact.
/plan-refactoring
Each recommendation includes the target (file, class, or function), the specific issues detected, an estimated effort level, and the expected impact on the repository score. This helps teams make data-driven decisions about which refactoring work to prioritize.
target-tests
Focus testing effort where it matters most. Uses mutation testing results, coverage data, defect prediction, and hotspot analysis to identify areas where additional tests would have the highest return on investment.
/target-tests
Returns a ranked list of files and functions that would benefit most from additional testing, along with the specific type of testing recommended (unit tests for uncovered branches, integration tests for high-coupling areas, etc.).
track-trends
Monitor codebase health over time. Runs score trend analysis and compares current metrics against historical baselines to identify whether quality is improving, stable, or degrading.
/track-trends
Shows the trajectory of the repository score and its components over recent history. Highlights significant changes (positive or negative) and correlates them with the commits that caused them.
report-debt
Generate a comprehensive technical debt report. Combines SATD detection, technical debt gradient, code smells, complexity analysis, and dependency issues into a structured debt inventory.
/report-debt
Produces a categorized inventory of all detected technical debt with severity ratings, locations, and estimated effort for remediation. Suitable for sprint planning, architecture reviews, or stakeholder reporting.
Usage Patterns
New Project Onboarding
When starting work on an unfamiliar codebase:
/setup-config-- generate configuration/onboard-codebase-- get the structural overview/audit-architecture-- understand dependency structure and problem areas
Pre-merge Review
Before merging a PR:
/focus-review-- prioritize which files to review/check-quality-- verify the quality gate passes/find-bugs-- check if changes touch high-risk areas
Sprint Planning
When deciding what technical work to prioritize:
/report-debt-- inventory current debt/plan-refactoring-- rank refactoring targets by impact/target-tests-- identify testing gaps
Periodic Health Checks
On a regular cadence (weekly, per-sprint, per-release):
/track-trends-- check health trajectory/audit-flags-- clean up stale flags/check-quality-- verify score is above threshold