The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands using a specialized CLI tool (listenhub). These commands are used for authenticating the user, managing configuration files (jq), and triggering the slide generation process. All commands follow a strict interactive flow requiring explicit user confirmation before execution.
[EXTERNAL_DOWNLOADS]: When optional narration is enabled, the skill uses curl to download generated audio files from listenhub.ai. This is a core functional requirement of the tool and targets the vendor's own verified infrastructure.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill reads and writes configuration data to a hidden directory (.listenhub/) and saves generated artifacts to the current working directory. It explicitly forbids saving to sensitive locations like ~/Downloads/. No unauthorized data transmission to third-party domains was detected.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data such as user-provided topics and content from external URLs. However, it implements a hard-gated interaction flow using the AskUserQuestion tool to confirm all parameters before processing, mitigating the risk of instructions embedded in the input data being automatically obeyed.

slides