The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill relies on npx atxp@latest to download and execute JavaScript code from the npm registry at runtime. This execution model allows for the acquisition and running of external code that can change independently of the skill's static instructions.
[EXTERNAL_DOWNLOADS]: Core functionality is dependent on the atxp CLI package and its dependencies, which are fetched from the public npm registry upon invocation.
[DATA_EXFILTRATION]: Provides high-privilege communication tools, including email send, email reply, and phone send-sms. These commands enable the transmission of data to arbitrary external recipients, posing an exfiltration risk for sensitive environment variables or local file content.
[PROMPT_INJECTION]: The skill ingests data from external untrusted sources such as web searches, X/Twitter posts, and inbound emails/SMS. This creates a significant surface for indirect prompt injection attacks, where malicious instructions embedded in external content could attempt to manipulate the agent's behavior.
[COMMAND_EXECUTION]: Implements all functionality via shell commands and subprocesses invoked through npx, giving the downloaded package access to the agent's execution environment.
[CREDENTIALS_UNSAFE]: The sensitive ATXP_CONNECTION token, which grants access to the agent's wallet and identity, is stored in a plain-text configuration file at ~/.atxp/config. The instructions note that this file uses export syntax, which could present a risk if handled improperly by shell interpreters.

atxp