Neon Soul 0.4.5

Automated soul synthesis for AI agents. Extracts identity from memory files, promotes recurring patterns to axioms (N>=3), generates SOUL.md with full proven...

This skill appears internally coherent, but it executes a bundled JavaScript CLI that will read your agent's memory files and produce SOUL.md — which means it will process potentially highly sensitive personal data. Before installing or scheduling it to run automatically: 1) Inspect scripts/neon-soul.mjs (or run it in a sandboxed/test workspace) to confirm there are no unexpected network calls or telemetry; 2) Run a dry-run first (use --dry-run) to see what would change and what it reads; 3) Back up your memory/ and existing SOUL.md before first run; 4) Ensure Ollama runs locally and is configured securely (the skill expects http://localhost:11434); 5) If you cannot audit the bundled code, prefer running the tool only in an isolated workspace or VM. These precautions reduce risk from hidden behavior in the compiled bundle.

Type: OpenClaw Skill Name: neon-soul-0-4-5 Version: 1.0.0 The skill is classified as suspicious primarily due to a path traversal vulnerability in `scripts/neon-soul.mjs`. The `xe` function, intended for path sanitization, allows access to any subdirectory within the user's home directory (`~`) and `/tmp`. While the skill processes only `.md` files, a malicious user could potentially direct the skill to read or overwrite sensitive `.md` files (e.g., crafted files in `~/.ssh` or `~/.aws`) by providing a controlled `--memory-path` or `--output-path` argument. Additionally, the skill executes external commands (`node`, `git`) via `execFileSync`, which, while implemented with array arguments for safety, still represents a potential RCE vector if underlying sanitization or command parsing were to fail. There is no evidence of intentional data exfiltration, backdoors, or malicious prompt injection within the skill's own instructions to the agent.