The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/export_config.sh utilizes eval to execute output generated by jq from the ~/.bytehouse_config.json file. While the values are escaped using @sh, the JSON keys are not sanitized before being passed to eval, creating a path for shell command injection if the configuration file is modified to include malicious keys.
[CREDENTIALS_UNSAFE]: The skill requires users to store sensitive information, specifically BYTEHOUSE_PASSWORD, in a local file at ~/.bytehouse_config.json. This file is then read by the execution scripts to authenticate against the API.
[DATA_EXFILTRATION]: The Python script scripts/run.py reads the user-provided BYTEHOUSE_PASSWORD and transmits it as a Bearer token in the Authorization header to the remote host defined by BYTEHOUSE_HOST. While this is the primary functionality for interacting with the ByteHouse service, it involves the transmission of credentials to an external endpoint.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted stream data from the ByteHouse Matrix Agent API.
Ingestion points: API response stream in scripts/run.py via response.iter_lines().
Boundary markers: None; the content is printed directly to the agent's output stream.
Capability inventory: The skill can perform network requests (requests.post) and read local configuration files.
Sanitization: None; JSON fields from the API (message['content']) are printed verbatim.

byted-bytehouse-smart-qa