The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions in SKILL.md explicitly recommend using eval "$(python scripts/xxx.py {params})" within a bash tool. This pattern is dangerous as it executes the output of a Python script directly in the shell context. If the Python script's output or the input parameters (like keyword or username) are manipulated, it could lead to arbitrary command execution on the host system.
[REMOTE_CODE_EXECUTION]: The Python scripts extract-profile-search.py and get-profile-details.py dynamically generate JavaScript strings that are intended for execution within a browser environment. While the scripts perform basic character escaping (replacing single quotes), the overall architecture of generating executable code from input parameters is a high-risk pattern.
[PERSISTENCE_MECHANISMS]: The skill attempts to maintain state across sessions by reading from and writing to a specific memory file located at {working-directory}/browser-act-skill-forge-memories/threads-scraper-threads-profile-search.memory.md. This is used to store 'experiences' and adjust strategies in future runs.
[DATA_EXFILTRATION]: The skill performs network requests to www.threads.com to extract profile information. While this is the stated purpose of the skill, the use of fetch() with custom headers (X-IG-App-ID) in get-profile-details.py demonstrates the ability to make arbitrary network calls from the browser context.

threads-profile-search