The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Reference to a malicious URL flagged by automated scanners. Evidence: Both SKILL.md and scripts/advanced_skill_processor.py reference https://docs.clawd.bot/tools/skills. Context: Automated security scans (URLite) have flagged this domain for botnet-related activity (Botnet|UR3BA220620CBB01B2-0200).
[PROMPT_INJECTION]: Presence of persona manipulation instructions and a significant indirect injection attack surface. Persona Manipulation: The 'System Prompt Integration' section in SKILL.md uses role-play prompts ('You are now an OpenClaw... expert') to override the agent's identity and instructions. Indirect Injection Surface: (1) Ingestion points: The skill's core research flow (Steps 2 and 3) requires the agent to browse and process content from untrusted sources such as ClawHub and GitHub. (2) Boundary markers: Absent. No instructions are provided to delineate untrusted content or to ignore potential instructions embedded within researched skills. (3) Capability inventory: The skill metadata requires python3 and bash capabilities, and the provided Python script imports the subprocess module. (4) Sanitization: Absent. The skill lacks any mechanism to validate or sanitize external data before fusing it into recommended code outputs.
[COMMAND_EXECUTION]: The skill requires environment permissions to execute local scripts. Evidence: The SKILL.md metadata explicitly requests python3 and bash binaries. Implementation: The skill includes a functional Python script scripts/advanced_skill_processor.py designed to be executed during the skill creation process.

advanced-skill-creator