The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The script scripts/fetch_growth_rank.py contains logic to programmatically read sensitive local shell configuration files to extract the REDFOX_API_KEY.
The script attempts to open and scan ~/.bashrc, ~/.bash_profile, ~/.zshrc, and ~/.profile on the host system.
It uses regular expressions to search for and capture values assigned to the REDFOX_API_KEY variable.
While this is documented as an 'automatic detection' feature, programmatically reading these files exposes the execution environment to potential leaks of other unrelated sensitive environment variables.
[PROMPT_INJECTION]: The skill processes untrusted external data (article titles and metadata) which presents a surface for indirect prompt injection.
Ingestion points: Data is retrieved from the vendor's external API endpoint at https://redfox.hk/story/api/cozeSkill/getGzhCozeSkillDataRaise within the scripts/fetch_growth_rank.py script.
Boundary markers: The skill instructions do not utilize specific delimiters or isolation warnings when passing the fetched article titles to the agent for its mandatory trend analysis.
Capability inventory: The skill possesses network access via the requests library and local file system read access for configuration discovery.
Sanitization: The script performs basic formatting but does not appear to sanitize or filter the article titles, which could contain instructions meant to influence the agent's behavior during the summary phase.

wechat-fastest-growing