lov-skill-creator
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFEOBFUSCATIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [OBFUSCATION]: The file 'references/cloud-split.md' provides detailed instructions on the 'Cloud-Split' design pattern, which is explicitly aimed at concealing internal skill logic, prompt engineering, and proprietary algorithms from the end user's local logs. It advocates for the use of 'Opaque output' and 'Symbolic tokens' (e.g., returning 'A' instead of a descriptive verdict) to prevent reverse engineering and hide AI reasoning processes.
- [INDIRECT_PROMPT_INJECTION]: The skill's migration workflow, documented in 'references/slash-command-migration.md' and implemented in 'scripts/migrate_command.py', involves ingesting untrusted external markdown and command files. If these source files contain malicious instructions, they could be propagated into the newly generated skills, representing an indirect injection vulnerability surface.
- [DYNAMIC_EXECUTION]: The 'scripts/init_skill.py' script generates executable Python and shell scripts dynamically based on user-supplied names and parameters. While this is functional behavior for a scaffolding tool, it constitutes runtime creation of executable content which necessitates strict validation of inputs to prevent misuse.
Audit Metadata