The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install the agent-browser utility from Vercel Labs' official GitHub repository. Vercel Labs is a trusted organization, and these downloads are part of the standard setup for the skill's research capabilities.\n- [COMMAND_EXECUTION]: The skill executes local shell scripts, specifically tool-discovery.sh (or tool-discovery.bat on Windows), to verify the presence of required tools such as Node.js, Python, and Git. It also utilizes the agent-browser CLI to perform automated research tasks like taking screenshots and extracting text from competitor websites.\n- [INDIRECT_PROMPT_INJECTION]: The skill possesses a potential attack surface for indirect prompt injection because it ingests untrusted data from external websites via agent-browser get text body and WebFetch.\n
Ingestion points: External pricing pages and competitor websites accessed during research sessions.\n
Boundary markers: The instructions do not define specific delimiters or warnings for the agent to ignore instructions embedded in the scraped content.\n
Capability inventory: The skill has the ability to read and write files, execute shell commands, and perform network operations.\n
Sanitization: There is no evidence of content sanitization or filtering applied to the data retrieved from external URLs.\n- [DATA_EXFILTRATION]: The skill provides advanced instructions for managing authenticated browser sessions for gated sites like LinkedIn or Twitter. This involves saving sensitive session tokens to local JSON files using agent-browser --state save. While the skill explicitly instructs users to add these files to .gitignore to prevent accidental exposure, the storage of session tokens in plaintext on the local filesystem is a security consideration.

paw-mkt-pricing