The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The instructions direct the agent to install a CLI tool by downloading a script and piping it directly to bash: curl -fsSL https://arc.afsd.io/install.sh | bash. This allows arbitrary remote code execution on the host system without prior content validation.
[EXTERNAL_DOWNLOADS]: The skill fetches an installation script from https://arc.afsd.io/install.sh. This domain is not identified as a trusted source or a standard domain for the vendor 'arcblock' in the analysis profile.
[COMMAND_EXECUTION]: The skill facilitates the execution of local shell commands like arc mcp and arc serve to bridge communication between clients and a local daemon.
[INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface by ingesting tool definitions and live mount data from external sources to populate the tool surface.
Ingestion points: Tool discovery via arc mcp and configuration via .mcp.json files.
Boundary markers: None identified; instructions do not provide delimiters for separating untrusted tool data from agent instructions.
Capability inventory: Execution of shell commands via the arc CLI and access to the AFS daemon.
Sanitization: No sanitization or validation logic is specified for data received through the MCP bridge or configuration files.

mcp