Openclaw Soul
ReviewAudited by ClawScan on May 18, 2026.
Overview
This skill appears to be a disclosed self-evolution framework, but it makes persistent changes to agent behavior, memory, permissions, and optional API credentials that deserve careful review before installation.
Install this only if you intentionally want a persistent self-evolving agent framework. Before proceeding, choose conservative permissions, review heartbeat/autonomous evolution settings, inspect credential handling, and check the bundled scripts and dependency skills that will be copied or installed.
Findings (6)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
The agent may continue updating memories and behavior over time, including during heartbeat runs, without the user reviewing every change.
The skill is not just installing files; it enables ongoing autonomous learning and memory updates after setup.
Fully autonomous daily learning and memory updates
Use supervised or advisory modes, review heartbeat settings, and disable autonomous evolution if you do not want ongoing background self-modification.
If elevated permissions are enabled, this and future agent actions may require fewer confirmations, increasing the impact of mistakes or unsafe instructions.
The deployment asks the user to raise OpenClaw's default permission level, including an elevated mode where most operations are automatically authorized.
建议提升权限等级... 设置为 elevated... 大部分操作自动授权
Prefer the standard permission level, avoid elevated unless you fully trust the environment, and restore stricter permissions after setup if desired.
API tokens for Moltbook or X/Twitter may be stored in plain shell startup files and become available to future shell sessions and local processes.
The optional EvoClaw setup can persist raw API keys into shell profile files, affecting the user's broader local environment.
If the human pastes a raw key: Save it automatically... echo "export MOLTBOOK_API_KEY='<the key they pasted>'" >> "$SHELL_PROFILE"
Use narrowly scoped tokens, prefer existing environment variables or a secret manager, and review shell profile changes before accepting credential setup.
Personal communication patterns and preferences may be stored and reused across sessions, which can be helpful but sensitive.
The skill intentionally observes user interaction patterns and persists them into memory/personality systems.
Dynamic Personality System: Observes user communication patterns over 10 conversations
Review what is written to USER.md, long-term memory, and memory directories; remove sensitive entries and avoid enabling broad retention if privacy is a concern.
Existing dependency skill installations could be replaced, and the final behavior depends partly on external skill packages or bundled fallback copies.
The documented preferred path installs or overwrites dependency skills from clawhub using --force.
clawhub install evoclaw --force clawhub install self-improving --force
Check which dependency skills will be installed or overwritten, review their sources, and avoid --force if you need to preserve existing local modifications.
If invoked, bundled scripts may run local shell commands, which can affect files or depend on local command behavior.
The static scan reports shell command execution in a bundled memory optimization script.
const result = execSync(
Inspect the memory optimization scripts before running them and avoid invoking them on untrusted workspaces or sensitive data until reviewed.