The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the bundled tts CLI located at /opt/hatch/bin/tts. This is the intended primary function of the skill and uses a specific, absolute system path.
[INDIRECT_PROMPT_INJECTION]: The skill processes arbitrary text supplied by users or read from script files, which creates a potential surface for indirect prompt injection.
Ingestion points: Data enters through the --text, --text-stdin, and --script parameters as documented in SKILL.md.
Boundary markers: The skill instructions correctly mandate the use of single-quoted heredocs (<<'TTS_INPUT') for single-shot synthesis. This is a critical security measure that prevents shell command injection during the synthesis process.
Capability inventory: The skill interacts with the /opt/hatch/bin/tts binary and reads local configuration files (voice_source.json, voices.json) using jq to select voice identifiers.
Sanitization: The instructions guide the agent to transform text into spoken form (e.g., spelling out URLs and abbreviations), though there is no automated sanitization for hidden instructions within the synthesized text itself. Given the output is audio, the impact of such injection is limited.

tts