The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external SaaS providers such as Jira, Slack, Discord, and Notion. This presents a surface for indirect prompt injection attacks where malicious instructions could be embedded in tickets, messages, or documents.
Ingestion points: External app actions involving providers like Email, Slack, Discord, Notion, Linear, Jira, and various CRMs (SKILL.md).
Boundary markers: The instructions utilize several boundary markers including connector_auth_boundary/v1 and a prepared-vs-observed boundary to separate internal logic from external data.
Capability inventory: The skill manages capabilities for external writes, sends, ticket mutations, and CRM updates.
Sanitization: The skill mitigates risks by requiring explicit connector_confirmation_gate/v1 checkpoints for any irreversible or externally visible actions.
[COMMAND_EXECUTION]: The skill includes a shell script snippet intended for the host platform's runtime environment.
Evidence: omh runtime record --skill connector-operator --harness connector-operator --status started in SKILL.md. This is a logging/status command used for tracking workflow execution within the Hermes framework.

omh-apps