The Agent Skills Directory

[Unverifiable Dependencies] (MEDIUM): The skill instructs the agent to install a global npm package (pinme) from an author ('glitternetwork') not on the trusted organizations list. This introduces a supply chain risk where malicious updates to the package could compromise the host environment.
[Command Execution] (MEDIUM): The skill relies on executing shell commands (npm install -g pinme, pinme upload) to perform its primary function. Executing commands from untrusted packages can lead to arbitrary code execution.
[Data Exposure] (LOW): The primary function is to upload data to IPFS, which is a public decentralized network. While the skill explicitly forbids uploading sensitive files like .env, .git, and node_modules, any content within the detected build directories (e.g., dist/, build/) will be made publicly available once uploaded.
[Indirect Prompt Injection] (LOW): The skill identifies and processes local directory structures which could be manipulated by an attacker to influence agent behavior.
Ingestion points: Local file system directory names and file structure.
Boundary markers: Absent. The skill does not use specific delimiters when passing directory paths to shell commands.
Capability inventory: Shell command execution (npm, pinme) and network access (via pinme upload).
Sanitization: Absent. There is no evidence of path sanitization before execution.

pinme