The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides instructions to clone a repository from an untrusted source (https://github.com/asz798838958/freeAgentIdentity.git) and execute its contents using Python and Node.js. This pattern is equivalent to a 'download and execute' attack vector if the repository content is malicious.
[EXTERNAL_DOWNLOADS]: The installation process fetches external scripts and dependencies from GitHub and public registries (NPM, PyPI) that are not within the control of the skill or a trusted organization.
[COMMAND_EXECUTION]: The skill requires the execution of multiple shell commands to set up the environment, install browser runtimes (playwright install chromium), and launch a web server (uvicorn main:app).
[INDIRECT_PROMPT_INJECTION]: The skill's architecture involves ingesting untrusted external data, such as email verification codes via IMAP/POP3 and captcha solutions from third-party providers. This creates an attack surface where malicious data could influence the agent's behavior during the automated registration process.
Ingestion points: email_provider.get_verification_code(), solve_captcha() results.
Boundary markers: None identified in the provided patterns.
Capability inventory: Subprocess calls (via Playwright), network operations (via httpx and FastAPI), and file writes (via SQLite/database operations).
Sanitization: Not explicitly documented; the skill interpolates verification codes directly into the browser automation flow.

abai-autoplus-chatgpt-registration