The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The setup instructions direct the user to download and execute a shell script from a remote GitHub repository using 'curl | bash'. While this repository is associated with the skill's author, the pattern of piping remote content directly to a shell is a significant security risk for arbitrary code execution.
[COMMAND_EXECUTION]: The documentation for setting up the skill in cloud environments includes instructions to use 'sudo' for installing system packages such as the OpenJDK and unzip utilities. This involves the execution of commands with elevated privileges.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it ingests and processes untrusted data from external sources. Ingestion points: The skill reads and summarizes content from GitHub pull request views, diffs, and API responses (documented in design/AGENT_PR.md). Boundary markers: There are no specific delimiters or instructions defined to isolate or ignore embedded malicious commands within the ingested text. Capability inventory: The agent has the ability to execute shell commands (Gradle, Git, gh), write to the file system, and perform network operations. Sanitization: The instructions do not specify any validation or sanitization of external data before it is incorporated into the agent's context.
[EXTERNAL_DOWNLOADS]: The skill performs several external downloads during setup and operation, including the CLI bootstrap script, release assets from GitHub, Gradle distributions, and Android SDK components.

compose-preview