The Agent Skills Directory

[COMMAND_EXECUTION]: The skill grants the agent the ability to execute a wide range of shell-level commands through the reflex utility.
Evidence: The instructions explicitly guide the agent to use commands like reflex xlsx, reflex rest, and reflex csv to perform direct library operations.
[REMOTE_CODE_EXECUTION]: The skill allows for the execution of arbitrary code strings generated or provided at runtime via script interpreters.
Evidence: The skill defines the reflex lua exec and reflex python exec commands specifically for running complex, dynamically generated logic.
[DATA_EXFILTRATION]: The skill provides both file system access and network capabilities, creating a direct path for data exfiltration.
Evidence: The agent can use libraries like Xlsx or Csv to read local data and the Rest library to send that data to external URLs via HTTP requests.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it is designed to ingest and process data from untrusted external sources.
Ingestion points: Data enters the agent's context through file reads (reflex xlsx open, reflex csv) and API responses (reflex rest get).
Boundary markers: The skill does not provide instructions for using delimiters or markers to differentiate between trusted instructions and untrusted data content.
Capability inventory: The skill provides arbitrary script execution (lua exec, python exec), file system modification (Fl.write), and network requests (rest).
Sanitization: There are no requirements or mechanisms described for sanitizing or validating data retrieved from external files or APIs before processing it.

reflex-scripting