How anshumanbh/vulnvibes works
VulnVibes accepts a GitHub pull request through its CLI or Python API, then runs threat modeling followed—when warranted—by skill-guided vulnerability investigation. The workflow uses GitHub tools to examine diffs and code, parses agent responses into structured findings, and can produce JSON and Markdown reports. The README documents the major workflow and capabilities; sampled source confirms key orchestration and tool relationships. Some integrations and report paths are documented rather than fully visible in the excerpts.
Main entry points: __init__.py PR Orchestrator · stages.py Stage Execution · cli.py CLI · definitions.py Agent Definitions
Components 17
Triage Workflow
- PR Orchestrator
vulnvibes/orchestrator/__init__.py - Stage Execution
vulnvibes/orchestrator/stages.py - Output Schemas
vulnvibes/orchestrator/schemas.py - Logging
vulnvibes/utils/logging.py
Analysis Resources
- Agent Definitions
vulnvibes/agents/definitions.py - Skill Registry
vulnvibes/skill_registry.py - Investigation Skills
vulnvibes/skills - Model Configuration
vulnvibes/config.py
External Access
- GitHub Tools
vulnvibes/providers/github_tools.py - GitHub Client
vulnvibes/providers/github_client.py
Results
- Response Parsers
vulnvibes/orchestrator/parsers.py - Report Generation
vulnvibes/orchestrator/reports.py
Other components
- Engineer
- GitHub
- Claude Agent SDK
Connections 24
- Engineer → CLI: runs analysis
README.md - Engineer → PR Input: provides PR
README.md - CLI → PR Orchestrator: invokes
vulnvibes/cli.py - CLI → PR Input: creates
vulnvibes/cli.py - PR Orchestrator → Stage Execution: runs stages
vulnvibes/orchestrator/__init__.py - PR Orchestrator → Agent Definitions: creates agents
vulnvibes/orchestrator/__init__.py - PR Orchestrator → GitHub Tools: creates tools
vulnvibes/orchestrator/__init__.py - PR Orchestrator → Report Generation: saves reports
vulnvibes/orchestrator/__init__.py - Stage Execution → Agent Definitions: queries agents
vulnvibes/orchestrator/stages.py - Stage Execution → Skill Registry: loads skill coverage
vulnvibes/orchestrator/stages.py - Agent Definitions → Skill Registry: builds skill prompts
vulnvibes/agents/definitions.py - Skill Registry → Investigation Skills: discovers skills
vulnvibes/skill_registry.py - Stage Execution → Output Schemas: uses schemas
vulnvibes/orchestrator/stages.py - Stage Execution → Response Parsers: parses responses
vulnvibes/orchestrator/stages.py - Stage Execution → GitHub Tools: requests GitHub data
README.md - GitHub Tools → GitHub Client: calls client
vulnvibes/providers/github_tools.py - GitHub Client → GitHub: fetches PR and code
README.md - GitHub Tools → Claude Agent SDK: registers MCP tools
vulnvibes/providers/github_tools.py - Stage Execution → Claude Agent SDK: queries agents
vulnvibes/orchestrator/stages.py - Agent Definitions → Model Configuration: selects models
vulnvibes/agents/definitions.py - CLI → Logging: configures logging
vulnvibes/cli.py - PR Orchestrator → PR Input: analyzes PR input
README.md - Response Parsers → Report Generation: supplies findings
- PR Orchestrator → Engineer: returns results
README.md