openclaw-persona-forge
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a bundled Python script (
gacha.py) to generate randomized character traits. The execution is handled through standard shell calls with user-supplied arguments that are sanitized and cast to integers within the script to prevent injection. - [INDIRECT_PROMPT_INJECTION]: The skill functions by ingesting user preferences to generate AI instruction files (
SOUL.md). This creates an attack surface for indirect prompt injection where malicious input could influence the behavior of agents using the generated persona files. - Ingestion points: User preferences provided during the guided design mode and identity tension forging steps (SKILL.md).
- Boundary markers: The output is formatted within structured Markdown blocks and includes a "concentration control" section designed to limit the persona's influence over core agent functions.
- Capability inventory: The skill utilizes file-writing tools to create
SOUL.mdandIDENTITY.mdand executes a local Python script for randomization. - Sanitization: The skill performs character filtering for filenames (replacing non-alphanumeric characters with hyphens) and validates the integer range for the randomization engine's arguments.
Audit Metadata