Nirvana Plugin

Local-first privacy-first inference. Your OpenClaw agent thinks locally and asks the cloud intelligently. Saves 85%+ tokens, protects privacy, agent learns f...

What to check before installing: - Review the code that reads SOUL.md, USER.md, MEMORY.md (src/context-stripper.js and privacy-auditor) to confirm stripping logic is correct and thorough. These files are very sensitive — the plugin needs to read them to strip data, but that requires trust. - Verify Docker image and model provenance before pulling (docker run ollama/ollama will pull from Docker Hub). If you need higher assurance, pull images from a verified registry or inspect image contents first. - The plugin caches cloud responses and writes audit logs to memory/*. Confirm audit logs are stored only locally at the configured path and are rotated/protected as appropriate. - The DELIVERY checklist includes uploading an archive to Google Drive — that is a publication step, not a runtime requirement. Do not upload any private data to external drives as part of setup. - If you enable cloud fallback, only add API keys you control and check that the context-stripping configuration (contextStripDepth, identityFilesNeverExport) is enforced before any cloud calls. Test with synthetic sensitive data to validate stripping. - Because the plugin will handle system prompts and can rewrite queries, do a small controlled test session and inspect outgoing HTTP payloads (or run in an air-gapped environment) to confirm no private fields are transmitted. Overall: the package looks coherent for its stated purpose, but it operates on very sensitive data and downloads large models — perform a code review and controlled testing before trusting it with real private files.

Type: OpenClaw Skill Name: project-nirvana-plugin Version: 1.0.0 Project Nirvana is a privacy-focused plugin designed to route AI queries to a local Ollama instance to reduce costs and protect sensitive data. The code implements a robust 'Context Stripper' (src/context-stripper.js) that redacts PII and sensitive files like SOUL.md and USER.md before any cloud fallback occurs. It includes comprehensive auditing (src/privacy-auditor.js) and metrics collection (src/metrics-collector.js) to ensure transparency. While it requests broad permissions to read sensitive agent files and execute Docker commands, these are strictly aligned with its stated purpose of local inference management and privacy enforcement. No evidence of malicious intent, data exfiltration, or unauthorized execution was found.