The Agent Skills Directory

[DATA_EXFILTRATION]: The skill transmits local audio and video files to Alibaba Cloud's Object Storage Service (OSS) for transcription. This behavior is the primary intended function of the skill and targets a well-known cloud service provider.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it ingests and processes untrusted transcription data from a remote API.
Ingestion points: Audio/video transcription text retrieved via the getTransResult API call in scripts/tingwu.py and scripts/poll_tasks.py.
Boundary markers: The skill does not currently implement specific boundary markers or 'ignore' instructions when formatting the transcription results into Markdown.
Capability inventory: The generated Markdown is intended to be consumed by other AI skills (e.g., funasr-transcribe/summary.py) for summarization and analysis.
Sanitization: There is no evidence of sanitization or filtering of the transcribed text content before it is interpolated into Markdown files.
[CREDENTIALS_UNSAFE]: The skill manages sensitive session cookies for authentication with Alibaba Cloud. It provides scripts (scripts/login.py, scripts/check_auth.py) and instructions for AI agents to capture and store these cookies locally in config/cookies.json. While this is a common pattern for automating web-based services, these cookies represent high-value targets for session hijacking.

tingwu-asr