balanced
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill manages user preferences by reading from and writing to a configuration file located at
~/.claude/skills/balanced/config.json. This is used to persist default modes and modifiers for the dialogue system. - [EXTERNAL_DOWNLOADS]: The skill requires the agent to perform web searches to validate Digital Object Identifiers (DOIs) associated with academic references. This ensures that factual claims are backed by verifiable evidence as part of the skill's primary function.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input as the primary subject of its analytical moves. (1) Ingestion points: User statements provided as arguments to the
/balancedcommand inSKILL.md. (2) Boundary markers: The instructions do not specify explicit delimiters or boundary markers for the user input. (3) Capability inventory: The skill has the ability to write to the local file system (config.json) and access the internet via a web search tool. (4) Sanitization: No input sanitization or filtering is defined before the agent processes the user statement. The associated risk is considered low given the intended analytical use case.
Audit Metadata