The Agent Skills Directory

[COMMAND_EXECUTION]: A command injection vulnerability exists in the cmd_headlines function within scripts/freshrss.sh. The --hours argument is interpolated into a subshell execution block $(date ...) without quoting or validation, enabling arbitrary shell command execution. An attacker could potentially gain control of the environment if the agent is persuaded to run this command with a malicious parameter.
[CREDENTIALS_UNSAFE]: The authentication mechanism in scripts/freshrss.sh sends the FRESHRSS_API_PASSWORD as a visible query parameter in a URL during the login process. This practice can lead to credential exposure in server logs, proxy logs, and command-line history.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8). It retrieves and presents data from external RSS feeds to the agent without sanitization or isolation.
Ingestion points: External RSS headlines and metadata are fetched using the FreshRSS API in scripts/freshrss.sh and outputted to the agent's context.
Boundary markers: No delimiters or specific instructions are used to separate untrusted feed content from the agent's system instructions.
Capability inventory: The skill possesses network access and local script execution capabilities.
Sanitization: The retrieved RSS data is displayed to the agent directly without any validation, escaping, or filtering.

freshrss