The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest untrusted text data and process it for audio synthesis. This creates a surface for indirect prompt injection where instructions could be embedded in the data to influence agent behavior or synthesis output.
Ingestion points: Data enters via the synthesize method in SecureTTSEngine (SKILL.md) and the filter method in TTSContentFilter (references/security-examples.md).
Boundary markers: The implementation lacks explicit boundary delimiters to isolate input text, though it mandates preprocessing steps and content filtering.
Capability inventory: The skill uses sf.write for file system writes, os.chmod for permission management, shutil.rmtree for deletion, and sounddevice for hardware audio output.
Sanitization: The skill implements several defensive measures, including regex filtering for sensitive patterns (passwords, API keys, tokens), length constraints (max 5000 characters), and character validation to strip shell-sensitive characters like ;, |, and &.

text-to-speech