The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from external social media platforms via the doubaoya.com API.
Ingestion points: The script scripts/fetch_trends.py retrieves titles and descriptions from Weibo, Douyin, and Bilibili.
Boundary markers: The skill's instructions do not define clear delimiters or specify that the agent should ignore instructions embedded in the fetched data.
Capability inventory: The agent is instructed to process this data to create 'exclusive cutting angles' and 'topic choices.'
Sanitization: No sanitization or filtering is applied to the retrieved content before the agent processes it.
Remote Control: The SKILL.md contains a specific instruction to pass a notice field from the API response directly to the user, providing a channel for the remote server to inject arbitrary messages into the agent's output.
[EXTERNAL_DOWNLOADS]: The skill script fetch_trends.py performs network operations to a non-whitelisted domain.
Evidence: The script makes POST requests to https://doubaoya.com/api/apis/trend/trending-hub-keyword/call to retrieve data.
[COMMAND_EXECUTION]: The documentation requires the user to execute local Python scripts and set environment variables.
Evidence: SKILL.md provides commands for export DOUBAOYA_API_KEY and running python3 $SKILL_PATH/scripts/fetch_trends.py.

trending-hub