The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content from local reference files and user-provided media, creating a surface for indirect prompt injection.
Ingestion points: The skill reads quote data from references/quotes.md and holiday metadata from references/holidays.md. It also utilizes user-supplied photos stored in the avatars/ directory for image generation prompts.
Boundary markers: There are no specific delimiters or instructions to ignore embedded commands when the text from quote files is interpolated into speech generation scripts or API payloads.
Capability inventory: The skill has access to media processing tools (ffmpeg), voice synthesis (sag), and network communication (curl) to send data to external APIs.
Sanitization: The shell script scripts/send-daily-quote.sh performs minimal escaping of text before including it in the JSON payload sent to the LINE API, which could lead to malformed requests if inputs contain special characters.
[EXTERNAL_DOWNLOADS]: The skill setup process involves downloading and installing several third-party command-line utilities.
The installation instructions use Homebrew to fetch sag (an ElevenLabs CLI), ffmpeg, and uv (a Python toolchain manager).
[COMMAND_EXECUTION]: The core functionality of the skill is driven by the execution of shell commands and external scripts.
The scripts/send-daily-quote.sh script coordinates the workflow by executing sag for text-to-speech, ffmpeg for audio conversion, and curl to interact with the LINE messaging API.
The SKILL.md instructions specify the use of uv run to execute image generation scripts and python3 for digital avatar video synthesis via HeyGen.

daily-voice-quote