config-reader
Warn
Audited by Gen Agent Trust Hub on Jun 14, 2026
Risk Level: MEDIUMDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill contains deceptive metadata. While the SKILL.md file explicitly states that the tool 'Does not modify any files', the config_reader.py file contains an 'update_config' function that overwrites files and a 'backup_config' function that creates new files on the system.
- [DATA_EXFILTRATION]: The skill is designed to read configuration files and return their full content. This behavior provides a vector for exposing sensitive information, such as environment variables, API keys, and database credentials, to the agent's context or an external party if the agent is instructed to exfiltrate the data.
- [PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection. It processes untrusted configuration files which may contain malicious instructions. If the agent treats the 'content' returned by 'read_config' as instructions, an attacker who can write to local configuration files could influence the agent's behavior.
- [COMMAND_EXECUTION]: The 'backup_config' function performs file system modifications and sets file permissions using 'os.chmod'. This contradicts the 'read-only' nature described in the skill's documentation.
Audit Metadata