The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install global software packages from public registries.
Installs the @bitbonsai/mcpvault package via npm install -g.
Installs the cli-anything-hub package via pip install.
[COMMAND_EXECUTION]: Automated shell commands are used to discover sensitive local data.
Executes a PowerShell command (Get-ChildItem) to recursively scan the user's home directory and OneDrive folders for Obsidian vaults.
Uses where.exe and which to verify executable locations on the host system.
[REMOTE_CODE_EXECUTION]: Configures the OpenCode agent to execute code provided by a third-party package.
Modifies opencode.json to add an MCP tool that runs @bitbonsai/mcpvault via npx.
[PROMPT_INJECTION]: The skill creates a vulnerability surface for indirect prompt injection.
Ingestion points: Processes all markdown (.md) files within the user's Obsidian vault via the installed MCP server.
Boundary markers: Absent. The instructions do not define delimiters or instruct the agent to ignore instructions embedded within the processed note data.
Capability inventory: The skill grants the agent read/write access to files and the ability to execute MCP-defined commands.
Sanitization: No filtering, escaping, or validation of note content is performed before processing.
[DATA_EXFILTRATION]: Requests broad access to personal data and modifies agent configuration files.
Grants the agent read/write access to the user's Obsidian vault, which typically contains private personal or professional information.
Modifies the agent's primary configuration file located at ~/.config/opencode/opencode.json.

opencode-obsidian