The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides explicit instructions for the agent to install a CLI tool using 'npx' or 'go install' from repositories associated with the author 'mvanhorn'. As this source is not part of the recognized trusted organizations list, executing these installation commands represents a risk of unverified remote code execution on the host machine.
[EXTERNAL_DOWNLOADS]: The skill initiates downloads of code and binaries from external package registries (npm) and source control platforms (GitHub) to set up the 'botsee-pp-cli' tool.
[DATA_EXFILTRATION]: The 'botsee-pp-cli' tool features a '--deliver webhook:' parameter. This capability allows the agent to route the results of any command—which may include sensitive account information, organization details, or API keys—to an external, attacker-controlled URL.
[COMMAND_EXECUTION]: The skill is designed to drive a local binary and instructs the agent to invoke shell commands with user-provided arguments, such as 'botsee-pp-cli ai-visibility-audit example.com'.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from untrusted external websites through its audit and search functionalities, creating a surface for indirect prompt injection. 1. Ingestion points: Ingests data from arbitrary URLs via the 'ai-visibility-audit' command and the local search index. 2. Boundary markers: No boundary markers or 'ignore' instructions are provided to the agent. 3. Capability inventory: The driven binary has network and file system access. 4. Sanitization: No sanitization of external content is specified before processing.

pp-botsee