The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill involves a workflow where an agent takes user-provided content (e.g., articles, briefs) and converts it into audio scripts. This creates a risk surface for indirect prompt injection if the source content contains malicious instructions designed to alter the agent's behavior.
Ingestion points: Untrusted data is processed when the agent is requested to create an audio version of a document as described in Section 1 of SKILL.md.
Boundary markers: The instructions do not prescribe the use of delimiters or protective framing to separate untrusted user input from the agent's internal production directives.
Capability inventory: The skill guides the use of Text-to-Speech (TTS) engines, audio processing, and RSS feed creation, which are standard for the use case but represent the capabilities potentially targeted by an injection.
Sanitization: The skill outlines normalization for audio clarity (e.g., phonetically spelling out symbols) in Section 3 and 4.2, but does not provide security-specific sanitization techniques.
[EXTERNAL_DOWNLOADS]: The skill links to and references guidelines and technical specifications from established organizations and service providers.
Evidence: References point to official documentation from Apple, Spotify, Google, Microsoft, YouTube, and the Audio Engineering Society (AES). These links are legitimate resources for the skill's context and follow safe practices for documentation.
[NO_CODE]: The skill contains only instructional text and evaluation metadata, with no accompanying scripts, executables, or code files.

podcast-production