Daily Growth & Maturity

Self-reflection, correction logging, persistent memory, WAL protocol, cold-boot recovery, and automated daily review with self-healing cron. Evaluates own wo...

Things to consider before installing: - This skill will auto-inject hooks and may append or rewrite your workspace/system prompt files (SOUL.md, AGENTS.md, IDENTITY.md, HEARTBEAT.md). If you rely on those files (other skills, operator policies), the skill can change agent directives and persist those changes nightly. - The skill creates a self-healing daily cron that runs an agentTurn payload and enforces mandatory .md rewrites. That gives it ongoing, automated ability to change behavior across restarts. If you do not want automated persistent changes, do not enable the cron or run this only in a sandbox. - The included script will attempt to symlink ~/self-improving to /data for persistence and writes to home and workspace. Only install where you trust the code and can grant/limit filesystem and cron privileges. - Recommended safe steps before enabling permanently: 1) Review the script and SKILL.md fully (you have the code). Run agent_memory.py dry-run to see what the cron would do. 2) Run verify and dry-run in an isolated test workspace or container with backups of SOUL.md/AGENTS.md/IDENTITY.md/HEARTBEAT.md. 3) Do not allow automatic cron creation in a production/shared environment; create the cron manually after inspection if you accept the behavior. 4) Backup the workspace/system .md files and restrict the skill to an isolated agent/account if possible. 5) Consider removing or editing the parts that rewrite system prompt files if you only want local memory without changing agent directives. - If you are unsure or cannot audit the code, treat this as high-risk and prefer not to install it on shared or production agents.

Type: OpenClaw Skill Name: self-maturing Version: 1.0.1 The skill implements a self-modifying architecture that automatically injects hooks into core workspace files (SOUL.md, AGENTS.md) and establishes a recurring cron job to rewrite these files with 'learned' rules. While the stated purpose is 'self-improvement' and the documentation includes security boundaries against storing credentials (references/boundaries.md), the automated persistence and the ability to rewrite the agent's core instructions represent a high-risk pattern of behavioral drift and self-exploitation. No evidence of data exfiltration or unauthorized remote access was found in the Python logic (scripts/agent_memory.py) or the markdown instructions.