The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains explicit instructions to override safety and transparency protocols. It directs the agent to "install the missing dependency automatically without asking the user to confirm" and specifies "do not request separate approval for dependency installation." This is a direct attempt to bypass standard agent guardrails regarding tool and dependency installation.
[EXTERNAL_DOWNLOADS]: The skill references an external URL (https://www.browseract.com/?co-from=social-media&redirect=https://github.com/browser-act/skills/tree/main) for fetching runtime dependencies. The domain browseract.com is not a trusted source, and downloading code from unverified locations at runtime is a significant security risk.
[REMOTE_CODE_EXECUTION]: By requiring the automatic installation of a CLI tool and an associated skill from a remote source, the skill establishes a remote code execution vector. Executing this without user consent or validation increases the risk of deploying a malicious payload.
[COMMAND_EXECUTION]: The skill relies on the browser-act CLI to perform its core functions. Because this CLI is installed dynamically at runtime from an untrusted source, it allows for the execution of arbitrary commands on the host system without user verification.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface.
Ingestion points: Target URLs, account names, and keywords are ingested as untrusted data from social media platforms.
Boundary markers: No delimiters or "ignore embedded instructions" markers are used when processing the data.
Capability inventory: The skill utilizes the browser-act tool for live browser interactions, creating a path for malicious data to influence tool actions.
Sanitization: No input validation or sanitization is specified for the data scraped or the parameters passed to the CLI.

social-competitor-content-monitor