The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions define an architecture where the agent ingests external data through platform context and storage modules.\n
Ingestion points: External data enters via fcode.context.parameters and fcode.storage.download as described in SKILL.md.\n
Boundary markers: While the documentation advises using try/catch for flow control, it does not explicitly instruct the agent to use specific delimiters or sanitization routines for input data.\n
Capability inventory: The defined environment includes capabilities for file system writes (fs.createWriteStream), network communication (via standard npm packages), inter-process execution (fcode.processes.run), and email dispatch (fcode.sendMail).\n
Sanitization: The skill notes the requirement to stringify data for the datastore but does not provide specific instructions for sanitizing external parameters before processing.\n- [DYNAMIC_EXECUTION]: The platform provides fcode.import() for module loading and fcode.processes.run() for executing external processes.\n
The documentation includes a safety constraint requiring that fcode.import() names be hardcoded string literals rather than dynamic variables.\n- [COMMAND_EXECUTION]: The skill describes standard file system operations using the Node.js fs module, specifically for handling temporary files within the designated process.env.TMP_DATA_DIR directory.\n- [EXTERNAL_DOWNLOADS]: The skill documents the @add-package directive for managing npm dependencies, using the well-known axios library as a standard example.

fcode-javascript