ralph talker

Self-improving conversational skill. Gets better at talking with every use. Saves what works to memory, evolves identity over time.

This skill is coherent with its description: it stores conversational memory and edits a small 'soul' file to evolve its persona. Before installing, decide whether you want a skill that persistently writes to the agent workspace. If you do, consider: (1) back up any existing SOUL.md or memory/ files, (2) run it in a sandbox or separate workspace first, (3) confirm the platform's permissions for skills to write files, and (4) monitor the memory/ directory and SOUL.md for unexpected edits. If you prefer no persistence, do not install or restrict file-write permissions. Note: the skill does not request network credentials or external services.

Type: OpenClaw Skill Name: ralph-talk Version: 1.0.0 The skill instructs the AI agent to write to and edit files (`memory/YYYY-MM-DD.md`, `SOUL.md`) based on user interaction. This creates a significant prompt injection vulnerability, as an attacker could potentially inject malicious commands or instructions into these files. If the agent later reads and executes content from these modified files without proper sanitization, it could lead to unauthorized actions, data exfiltration, or remote code execution. While the stated purpose of self-improvement is benign, the method introduces a high-risk attack surface.