The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The instructions in Step 1 (Dimensions 1 and 2) explicitly direct the agent to 'actually run' installation commands (such as npm ci or pip install -r) and test scripts found within the target repository. Since the repository to be assessed is provided as a user-controlled input ($1), a malicious user can provide a repository containing harmful scripts (e.g., via preinstall hooks in package.json) that will be executed by the agent.
[COMMAND_EXECUTION]: The skill uses the Bash tool to perform multiple sensitive operations, including making network requests via curl to a gateway URL ($PYLOT_GATEWAY_URL) to fetch authentication tokens, and executing git clone using those tokens. These operations handle sensitive credentials ($PYLOT_BROKER_TOKEN, $PYLOT_DISPATCH_TOKEN) and involve dynamic command construction based on external variables.

agent-readiness