The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill uses bunx ctx7@latest to download and execute the latest version of the ctx7 package from the NPM registry. This involves executing code from a third-party source that is not verified as a trusted vendor, potentially allowing for the execution of malicious scripts if the package is compromised.
[EXTERNAL_DOWNLOADS]: The scripts/llms-probe.sh utility performs network requests to arbitrary user-provided domains using curl. It specifically uses the -k (or --insecure) flag, which disables SSL certificate verification, making the connection susceptible to man-in-the-middle attacks.
[COMMAND_EXECUTION]: The skill relies on shell script execution (bash scripts/llms-probe.sh) and external CLI tools (bunx). These execution paths could be exploited if user-supplied inputs are not properly sanitized before being passed as arguments.
[INDIRECT_PROMPT_INJECTION]: The skill ingests documentation data from external websites, creating a surface for indirect prompt injection.
Ingestion points: The agent fetches documentation files like llms.txt and llms-full.txt from remote servers and processes output from the ctx7 tool.
Boundary markers: The instructions lack specific markers or delimiters to help the agent distinguish between its instructions and the data fetched from documentation.
Capability inventory: The environment allows for file system access, network requests via curl, and package execution via bunx.
Sanitization: There is no evidence of filtering, escaping, or validating the remote documentation content before it is presented to the agent.

lib-docs