The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to process and execute content from project files, which creates an attack surface for indirect prompt injection if those files originate from untrusted sources.\n
Ingestion points: The agent is guided to read and process contract sources in contracts/, migration scripts in migrations/, and test files in test/ (refer to core-compile.md, core-migrations.md, and core-testing.md).\n
Boundary markers: There are no explicit boundary markers or warnings defined to distinguish between trusted instructions and potentially malicious data within these project files.\n
Capability inventory: The tronbox migrate and tronbox test commands execute arbitrary JavaScript, while tronbox compile processes Solidity code, all of which could be manipulated by malicious inputs.\n
Sanitization: No sanitization or validation of the content within the project files is mentioned before execution by the tool.\n- [REMOTE_CODE_EXECUTION]: Several documented CLI commands involve fetching and executing code from external or remote sources.\n
The tronbox unbox command allows for the download and extraction of project templates from remote URLs or registries, which could lead to the execution of untrusted code if the source is compromised (features-init-unbox.md).\n
The tronbox init command automatically executes npm install when a package.json file is present in a sample project, potentially installing and running scripts from third-party packages (features-init-unbox.md).\n- [DYNAMIC_EXECUTION]: The migration system uses dynamic execution to manage the deployment of smart contracts.\n
Migration scripts (JavaScript) are executed within a virtual machine (VM) context, enabling the execution of logic defined within the migrations/ directory at runtime (core-migrations.md).\n- [CREDENTIALS_UNSAFE]: The documentation addresses the handling of private keys for network deployment.\n
While example placeholders are used, the documentation correctly identifies and promotes the security best practice of using environment variables (e.g., process.env.PRIVATE_KEY) and .env files to manage secrets securely (core-config.md, core-cli.md).

tronbox