The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates workflows that ingest untrusted external data, such as documents or user-provided topics, into LLM prompts (e.g., openrouter/claude-sonnet-45) to generate podcast scripts. This creates a surface for instructions embedded in source documents to influence the agent's behavior.
Ingestion points: SKILL.md (via placeholders like <your-document-content> and <podcast-script>).
Boundary markers: Absent; external data is interpolated directly into prompts without delimiters or warnings to ignore embedded instructions.
Capability inventory: The skill uses the Bash tool to execute belt CLI commands for audio generation, TTS, and media merging.
Sanitization: Absent; there is no evidence of input validation or escaping for the processed content.
[EXTERNAL_DOWNLOADS]: The skill refers to external installation scripts and CLI tools from the inference-sh organization on GitHub.
Evidence: https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md.
Context: These references are used for standard setup of the required belt CLI and originate from a trusted source.
[COMMAND_EXECUTION]: The skill relies on the Bash tool to interact with the belt CLI for all audio processing and AI model invocations.

ai-podcast-creation