The Agent Skills Directory

[CREDENTIALS_UNSAFE] (HIGH): The skill accesses and clears sensitive information from the system keyring (service: aigne-doc-smith-publish) and local configuration files (~/.aigne/doc-smith-connected.yaml). Unauthorized or accidental invocation could result in the loss of critical authentication tokens.
[PROMPT_INJECTION] (HIGH): The skill is highly vulnerable to Indirect Prompt Injection. It processes site hostnames derived from user input or context and passes them directly to shell-executed scripts.
Ingestion points: Site hostnames extracted from the conversation or user messages (e.g., in the clear-auth.mjs command).
Boundary markers: None are present in the instructions to prevent the agent from being misled by malicious site names (e.g., a name containing shell metacharacters).
Capability inventory: Executes local Node.js scripts via child_process (implied by the node command) that modify system-wide and local security settings.
Sanitization: The skill description lacks any requirement or evidence of input validation, making it susceptible to command injection if a malicious hostname is provided.
[COMMAND_EXECUTION] (MEDIUM): The skill uses the node command to execute local maintenance scripts with dynamic arguments. This presents a risk of argument injection if the underlying scripts do not strictly validate the provided site hostnames.

doc-smith-clear