The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources and user prompts which could potentially contain malicious instructions designed to manipulate the agent's execution flow.
Ingestion points: The skill accepts external image URLs for identity training (Step 1) and user-supplied text for script generation (Step 4) across SKILL.md.
Boundary markers: The instructions lack specific delimiters or "ignore embedded instructions" warnings to isolate external data from the agent's core logic.
Capability inventory: The skill is granted Bash, Write, Read, and Agent tools, enabling it to execute system commands and modify files.
Sanitization: There is no documentation of sanitization, validation, or filtering of the content fetched from external URLs before processing.
[COMMAND_EXECUTION]: The skill uses the Bash tool to perform multiple operations via a command-line interface.
Evidence: The workflow documents several infsh app run commands (e.g., phota/train, inworld/text-to-speech-2, pruna/p-video-avatar) that handle external parameters and file paths.
[EXTERNAL_DOWNLOADS]: The skill relies on multiple external AI service modules to process data and generate media assets.
Evidence: It integrates tools from providers including pruna, phota, and inworld. References to google services are recognized as originating from a well-known and established service provider.

ai-podcast