SoulForge

Automatically evolves your SOUL.md by analyzing your conversation patterns and behavior, then proposes updates to reflect who you truly are over time.

This skill appears to do what it says: it processes local session logs and proposes edits to SOUL.md. Before installing, review the scripts (observe.py, reflect.py, forge.py) yourself because they will read your session history (which may contain private or sensitive data) and store aggregated observations in memory/observations.json and backups under memory/backups. If you want to review everything before any changes, keep auto-observation disabled (add soulforge: observe: false to your config) and avoid running forge with any auto-accept flags. If you have sensitive data in session logs you don't want processed, remove or redact those logs first. Finally, consider inspecting/validating the installed scripts (and any CLI flags like --auto-accept) to ensure they behave as you expect.

Type: OpenClaw Skill Name: soul-forge Version: 1.0.0 The skill bundle is classified as suspicious due to a potential path traversal vulnerability in `forge.py`, `observe.py`, and `reflect.py`. These scripts accept `--soul` or `--file` arguments and use the provided paths directly with `Path()` for file operations (read/write) without explicit sanitization. While the documentation and security manifests consistently claim local-only, read-mostly, and user-approved operations, a malicious input path (e.g., `../../../../etc/passwd`) could allow unauthorized file access. There is no evidence of intentional malicious behavior such as data exfiltration, backdoors, or prompt injection designed to subvert the agent's core directives; the stated purpose and code logic are aligned with self-reflection and evolution of the agent's `SOUL.md` file.