The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill's primary execution method is npx xactions-mcp, which downloads and executes code from the npm registry. The author 'nichxbt' and the package 'xactions-mcp' are not verified trusted entities, representing the execution of third-party code from an unknown source.
[COMMAND_EXECUTION]: The setup instructions provide CLI commands such as xactions mcp-config --client [client] --write. These commands instruct the agent to execute binary code that modifies local system configuration files.
[INDIRECT_PROMPT_INJECTION]: The skill provides numerous tools for scraping untrusted external data from X/Twitter, creating a vulnerability surface where malicious external content could influence agent behavior.
Ingestion points: x_get_tweets, x_search_tweets, x_get_thread, x_get_profile, x_get_explore, and x_get_trends in SKILL.md.
Boundary markers: The skill descriptions do not specify any delimiters or instructions for the agent to ignore embedded commands in the scraped data.
Capability inventory: The skill grants the agent high-impact capabilities including x_post_tweet, x_follow, x_send_dm, x_delete_tweet, x_update_profile, and x_reply.
Sanitization: No sanitization or validation of the ingested external content is mentioned.
[DATA_EXFILTRATION]: The skill handles sensitive authentication data via the XACTIONS_SESSION_COOKIE and provides tools designed to extract private data from the user's account, such as x_export_account, x_export_dms, and x_get_bookmarks.

xactions-mcp-server