The Agent Skills Directory

PROMPT_INJECTION (LOW): Indirect Prompt Injection vulnerability surface. The skill ingests untrusted data from external Mattermost chat conversations that could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: Output from the mm dms and mm channels commands.
Boundary markers: Absent. The skill does not define specific delimiters or instructions for the agent to treat chat content as untrusted data.
Capability inventory: The agent has the ability to read private messages, list channels, and use shell redirection.
Sanitization: The skill claims to redact secrets but does not provide sanitization against natural language command injection within chat messages.
DATA_EXFILTRATION (LOW): Local data exposure risks.
The skill accesses its own configuration file at ~/.config/mattermost-cli/config.toml which contains sensitive API tokens (MM_TOKEN). While necessary for the tool's function, this is a sensitive credential path.
Example workflows suggest writing sensitive JSON chat data to /tmp/bob-chat.json. On multi-user systems, data written to /tmp/ without specific permissions may be accessible to other local users.
COMMAND_EXECUTION (SAFE): The skill utilizes a local mm CLI tool. There are no patterns of remote code execution, dynamic code generation, or downloads from untrusted sources detected in the provided file.

mattermost-cli