The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The file SKILL.md contains a hardcoded FEISHU_RECEIVER_OPEN_ID (ou_18b8063b232cbdec73ea1541dfb74890) associated with a specific user. While not a secret key, hardcoding unique user identifiers is a poor security practice.
[INDIRECT_PROMPT_INJECTION]: The skill's workflow involves reading 30-50 news articles from a cache to generate podcast scripts. This untrusted external data could potentially contain malicious instructions intended to influence the AI's script generation or subsequent actions.
Ingestion points: External news content read by scripts/daily_generate.py.
Capability inventory: The skill executes local Python scripts and shell commands, and performs network operations to RSS and notification services.
Boundary markers: None identified in the provided instructions to differentiate between news content and agent instructions.
Sanitization: No evidence of sanitization or filtering of the ingested news articles is present in the provided files.
[COMMAND_EXECUTION]: The skill relies heavily on executing local scripts (e.g., publish_to_rss.py, notify_feishu.py, daily_generate.py) and shell commands. The publish.sh script dynamically exports environment variables from a .env file using xargs, which could lead to command injection if the .env file is compromised.

podcast-publish-xiaoyuzhou