The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted audio content from external URLs or local files and extracts text transcripts.
Ingestion points: Audio data is ingested via the --audio argument in scripts/transcribe_audio.py.
Boundary markers: None identified in the output formatting for transcripts.
Capability inventory: The skill uses urllib.request for network operations and pathlib for file system writes.
Sanitization: No explicit sanitization or filtering is performed on the transcribed text before it is returned or stored.
[DATA_EXPOSURE]: The helper script scripts/transcribe_audio.py accesses sensitive configuration files and environment variables.
Evidence: The script reads from ~/.alibabacloud/credentials and .env files to retrieve the DASHSCOPE_API_KEY.
Context: This is standard behavior for an Alibaba Cloud provider skill and the data is sent only to the official dashscope.aliyuncs.com domain.
[COMMAND_EXECUTION]: The skill instructions and validation steps involve executing shell commands and Python scripts.
Evidence: SKILL.md contains commands for py_compile, curl, and executing the bundled scripts/transcribe_audio.py.
Context: These commands are used for installation, validation, and interacting with the documented service API.

aliyun-qwen-asr