The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The SKILL.md file suggests an installation method using curl -sL https://skills.sh/paperclip | bash. This pattern downloads and executes a script from a remote server directly in the shell.
[EXTERNAL_DOWNLOADS]: The skill fetches the paperclip-cli package from npm and provides instructions to clone the author's official GitHub repository at github.com/ckorhonen/paperclip-cli.git.
[DATA_EXFILTRATION]: Provided scripts such as comment.sh and issues.sh access the PAPERCLIP_API_KEY environment variable and include it in network requests to the endpoint defined by PAPERCLIP_API_URL.
[PROMPT_INJECTION]: The skill processes potentially untrusted data from an external API, creating a surface for indirect prompt injection.
Ingestion points: scripts/issues.sh, scripts/status.sh, and scripts/comment.sh retrieve issue data and comments from the API.
Boundary markers: No delimiters or safety warnings are present when displaying retrieved content.
Capability inventory: The skill performs network operations via curl and executes shell commands via the paperclip CLI.
Sanitization: No sanitization or escaping is applied to the data fetched from the API.
[COMMAND_EXECUTION]: Utility scripts in the scripts/ directory perform shell operations and use python3 -c for processing data at runtime.
[CREDENTIALS_UNSAFE]: The documentation includes instructions to store sensitive API credentials in shell configuration files like ~/.zshrc or ~/.bashrc.

paperclip