Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
Conversation Recovery
v1.0.0Capture and recover conversation state across OpenClaw sessions. Use when conversations get interrupted, span multiple sessions, or need context restoration....
⭐ 0· 124·0 current·0 all-time
byhaidong@harrylabsj
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
medium confidencePurpose & Capability
The name/description align with the implementation: session/snapshot models, capture/recover APIs, local JSON storage, and analysis/compression helpers. However the registry metadata lists no required binaries or env vars while SKILL.md expects npm/node (npm install, build) and the code reads process.env.CONVERSATION_RECOVERY_STORAGE and process.env.HOME. The skill should declare Node/npm as required binaries and document the optional CONVERSATION_RECOVERY_STORAGE env var in the registry metadata.
Instruction Scope
SKILL.md and the exported API focus on capturing/recovering conversation state and managing snapshots; the implementation reads/writes JSON files under a dedicated directory (~/.openclaw/conversation-recovery by default). There are no instructions or code paths that reach outside these responsibilities (no network endpoints, no scanning of unrelated system files). The analyzer notes LLM-based extraction would be a production-time call but currently contains a simulated implementation — no active external LLM calls were found in the shipped code.
Install Mechanism
There is no packaged install spec in the registry, but SKILL.md requires running npm install and npm run build; package.json and package-lock.json are included. Installing will pull dependencies from the npm registry and write node_modules to disk. This is a standard Node.js build workflow but the registry should advertise node/npm requirements; review package.json dependencies before running npm install to confirm no unexpected packages.
Credentials
The skill requests no credentials (good) but reads/writes files under the user's home directory (default location ~/.openclaw/conversation-recovery) and uses the CONVERSATION_RECOVERY_STORAGE env var to override storage. The registry's required env list is empty despite the SKILL.md and code referencing that environment variable. File write/delete operations (createSnapshot, deleteSnapshot, deleteSession) are present — these are appropriate for the feature but are destructive. The lack of declared env/config requirements in the registry is an inconsistency that reduces transparency.
Persistence & Privilege
The skill does not request 'always: true' or other elevated platform privileges. It can be invoked autonomously per platform defaults (disable-model-invocation is false), which is normal for skills. Its persistence is limited to storing JSON files in a per-user directory; it does not modify other skills' configurations or system-wide settings.
Assessment
What to check before installing:
- Node/npm requirement: SKILL.md requires running 'npm install' and 'npm run build' but the registry metadata lists no required binaries — ensure you have Node.js and npm and be comfortable running them.
- Review package.json/package-lock.json for dependencies before running npm install to spot any unexpected packages.
- Storage location: by default the skill writes JSON files to ~/.openclaw/conversation-recovery (sessions/ and snapshots/). If you prefer isolation, set CONVERSATION_RECOVERY_STORAGE to a custom path (export CONVERSATION_RECOVERY_STORAGE=/path) before use or run inside a sandbox/container.
- Data deletion: functions exist to delete snapshots and sessions permanently. Back up any important conversation data before using deletion/cleanup APIs.
- Network/LLM: current code simulates LLM extraction and does not perform external network calls, but roadmap mentions future LLM integrations — if you upgrade later, re-check for outbound network calls and required API credentials.
- Install in a controlled environment (VM/container) if you want to inspect runtime behavior first. If you lack the ability to audit dependencies, avoid running 'npm install' as root or on a production machine.Like a lobster shell, security has layers — review code before you run it.
latestvk976j9d7tat010j2rkrz2csk3h8384xf
License
MIT-0
Free to use, modify, and redistribute. No attribution required.