How pdmiv/harp works
HARP is an authorized-lab penetration-testing prototype. Operators configure a target and runtime in the dashboard or connect Codex through MCP; the runner performs reconnaissance, coordinates oracle planning and proxy actions, validates and executes commands in Kali, and retains evidence and outcomes. Additional branches support Whippet AD analysis, bounded CVE/web research, replay, and settings snapshots. The graph follows README workflows and shown import/reference evidence; several subsystem internals are unsampled, so uncertain links are omitted.
Main entry points: harp.py Lab runner · harp_web.py HARP dashboard · harp_mcp.py Codex MCP · harp_whippet.py Whippet AD analysis
Components 21
Operator interfaces
- HARP dashboard
harp_web.py - Codex MCP
harp_mcp.py
Session orchestration
- Lab runner
harp.py - Deterministic recon
harp.py - Command validation
harp.py - Kali runtime
Evidence and research
- Capability evidence
harp_evidence.py - Planning checks
harp_strategy.py - CVE and web research
harp_research.py - Whippet AD analysis
harp_whippet.py - Whippet batch queries
harp_whippet_batch.py
Model integrations
- Oracle planning
harp_nim.py - Proxy model
harp.py - OpenAI transport
harp_openai.py - Z.AI transport
harp_zai.py - NVIDIA NIM transport
harp_nim.py
Session records
- Flags and artifacts
harp_artifacts.py - Session replay
harp_replay.py - Settings snapshots
harp_snapshots.py - Session archive
harp_session_archive.py
Other components
- Operator
Connections 24
- Operator → HARP dashboard: configures and starts
README.md - Operator → Codex MCP: connects via
README.md - HARP dashboard → Lab runner: starts sessions
harp_web.py - Codex MCP → Lab runner: uses session tools
harp_mcp.py - Lab runner → Deterministic recon: performs
README.md - Lab runner → Oracle planning: requests plans
README.md - Lab runner → Proxy model: delegates objectives
README.md - Lab runner → Command validation: validates commands
README.md - Command validation → Kali runtime: permits execution
README.md - Lab runner → OpenAI transport: calls when selected
harp.py - Lab runner → Z.AI transport: calls when selected
harp.py - Lab runner → NVIDIA NIM transport: calls when selected
harp.py - Lab runner → Capability evidence: updates evidence
harp.py - Lab runner → Planning checks: uses planning checks
harp.py - Lab runner → CVE and web research: uses research
harp.py - Lab runner → Whippet AD analysis: queries AD findings
harp.py - Lab runner → Whippet batch queries: uses managed queries
README.md - Lab runner → Flags and artifacts: records artifacts
harp.py - Codex MCP → Whippet AD analysis: queries findings
harp_mcp.py - Codex MCP → Flags and artifacts: records evidence
harp_mcp.py - HARP dashboard → Session replay: offers replay
harp_web.py - HARP dashboard → Settings snapshots: manages settings snapshots
harp_web.py - HARP dashboard → Session archive: uses session archives
harp_web.py - Lab runner → Session archive: stores session history