The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill contains instructions for the immediate execution of remote scripts from unverified sources via piped shell commands. Specifically, it includes curl -fsSL https://pi.dev/install.sh | sh and curl -fsSL https://opencode.ai/install | bash. These patterns are dangerous as they bypass manual review of the script contents before execution on the host machine.
[COMMAND_EXECUTION]: The skill recommends global installation of Node.js packages from unverified authors, such as @earendil-works/pi-coding-agent and @ampcode/cli, which presents a supply-chain risk if these packages are malicious or compromised.
[PROMPT_INJECTION]: The skill promotes a workflow where AI agents automatically process instructions from files within a project repository (e.g., AGENTS.md, CLAUDE.md, and .mcp.json). This creates a vulnerability to indirect prompt injection, where an attacker could influence the agent's behavior by placing malicious instructions in these files.
Ingestion points: Tools are instructed to read AGENTS.md, CLAUDE.md, .mcp.json, and .claude/rules/ from the local workspace.
Boundary markers: None identified; the agents are explicitly intended to follow instructions contained in these files.
Capability inventory: The agents configured by this skill typically possess broad read/write/execute permissions on the user's machine.
Sanitization: The skill does not implement or suggest any sanitization for the external content before it is processed by the agents.

ai-dev-tools