cost
Pass
Audited by Gen Agent Trust Hub on Jun 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill manifest instructs the agent to run a local Python script (
cost.py) to process data. This script uses only standard libraries and does not include any network-related code. The manifest contains a hardcoded path for a specific user environment, which is a portability issue rather than a security concern.- [DATA_EXFILTRATION]: The skill reads conversation history from~/.claude/projects/. This is the intended behavior to calculate costs. There are no network operations (such as curl, wget, or requests) present in the code, ensuring that the processed data remains on the local machine.- [PROMPT_INJECTION]: The skill processes untrusted historical conversation data. The risk of indirect prompt injection is mitigated because the script only extracts numeric metadata and model identifiers using standard JSON parsing and does not treat message content as executable instructions.
Audit Metadata