The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill creates a vulnerability surface by ingesting untrusted data from external sources.
Ingestion points: Data enters the agent context via messages_read and attachments_fetch tools, which pull content from Telegram, Discord, and Slack.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore or isolate instructions embedded within the ingested messages.
Capability inventory: The skill possesses the ability to send messages (messages_send) and schedule recurring tasks (hermes cron), which could be abused if the agent is influenced by injected content.
Sanitization: No evidence of sanitization, filtering, or validation of the external content is present in the skill instructions.
[PERSISTENCE]: The skill documents the use of hermes cron to schedule recurring tasks. While this is presented as a primary feature for automation and periodic check-ins, it utilizes a standard persistence mechanism to maintain agent activity across sessions.
[DYNAMIC_EXECUTION]: The skill describes a "closed learning loop" where the agent "creates skills from experience" and "improves them during use." This involves the dynamic generation and modification of executable skills/prompts at runtime.
[COMMAND_EXECUTION]: The skill provides instructions for executing shell commands to manage the Hermes agent, including activating virtual environments, checking status, and adding cron tasks (e.g., hermes status, hermes chat).

hermes-integration