The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The installation process involves cloning a repository and executing a shell script (hermes-kanban-install.sh) from a third-party GitHub repository (GumbyEnder/hermes-kanban). This introduces a risk of executing unverified code during the setup phase.
[PRIVILEGE_ESCALATION]: The skill instructions direct the user to modify system-level security configurations using the netsh advfirewall command. This opens a network port (27124) to allow inbound traffic, which could expose the local Obsidian vault to the network, especially if the 'Trust mode' is set to 'auto'.
[EXTERNAL_DOWNLOADS]: The skill downloads plugin binaries and assets (main.js, manifest.json, styles.css) from the obsidian-community/obsidian-kanban repository. While this is a well-known community organization, it represents an external dependency fetched at runtime.
[COMMAND_EXECUTION]: The skill relies on various shell commands for its setup and operation, including npm install, npm run build, and git clone, which are executed in the local environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from Obsidian Markdown files which can be influenced by third parties or other plugins.
Ingestion points: Kanban card titles, descriptions, and metadata are read from Markdown files via the REST API endpoints.
Boundary markers: None identified in the skill instructions; the content is processed as standard Markdown lists.
Capability inventory: The agent has the ability to read, create, and modify cards, as well as generate summaries of board states.
Sanitization: No evidence of sanitization or filtering of Markdown content before it is processed by the agent.

hermes-kanban-obsidian-integration