The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides templates for building CLI applications that ingest external data and possess powerful system capabilities. This structure is susceptible to indirect prompt injection if user-controlled input is processed without strict boundaries in an agentic environment.\n
Ingestion points: Command-line arguments (Args), flags (Flags), and interactive text input components (TextInput) across examples/core.md and examples/advanced.md.\n
Boundary markers: The provided examples do not demonstrate the use of delimiters or 'ignore' instructions for untrusted input.\n
Capability inventory: File system writing (the writeConfig function in examples/core.md), network communication (the fetch implementation in examples/advanced.md), and shell command execution (execa usage in examples/testing.md).\n
Sanitization: While basic validation is shown for numeric inputs (e.g., the parse function in examples/core.md), there are no examples of sanitizing string inputs against injection attacks.\n- [REMOTE_CODE_EXECUTION]: The skill documents the use of @oclif/plugin-plugins in examples/advanced.md, which enables the runtime installation and execution of external npm packages as CLI plugins. This is a core feature of the framework but allows for the execution of code from external sources.\n- [COMMAND_EXECUTION]: Patterns in examples/testing.md utilize the execa library to spawn external processes, such as executing git clone for data synchronization. This is a legitimate functional requirement for many CLI tools.\n- [EXTERNAL_DOWNLOADS]: Code examples in examples/advanced.md (Pattern 7) and examples/core.md (Pattern 10) demonstrate the use of the fetch API to download content from remote URLs and send telemetry data to a placeholder endpoint.

cli-framework-oclif-ink