The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill's documentation and bootstrap scripts (scripts/bootstrap.sh, README.md) recommend installing components by piping a remote script to a shell: curl -fsSL https://raw.githubusercontent.com/awp-worknet/ardi-skill/main/install.sh | sh. This script is hosted on the author's GitHub repository and executes with user privileges.\n- [EXTERNAL_DOWNLOADS]: The skill downloads the ardi-agent executable from GitHub releases during the installation process. Additionally, an automated scan flagged the public RPC endpoint https://base.meowrpc.com (referenced in src/rpc.rs) as malicious; however, this is a well-known public service provider.\n- [COMMAND_EXECUTION]: The skill installs a systemd timer and service (ardi-mine.timer, ardi-mine.service) to run mining tasks in the background autonomously. It also executes the ardi-agent CLI and is designed to spawn other agent frameworks such as Claude Code or Hermes to perform riddle-solving tasks.\n- [PROMPT_INJECTION]: The skill ingests riddle content from an external coordinator API (api.ardinals.com), which presents a surface for indirect prompt injection. \n
Ingestion points: Riddle data is fetched and processed in src/cmd/context.rs. \n
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the prompt templates (e.g., tools/auto-mine/prompt/ardi-mine-tick.md). \n
Capability inventory: The skill can execute shell commands (src/main.rs) and perform blockchain transactions through awp-wallet (src/tx.rs). \n
Sanitization: External riddle content is not sanitized before being presented to the agent for reasoning.

ardi