The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill provides templates and instructions for users to input sensitive Bilibili session cookies (sessdata, bili_jct, buvid3) directly into the agent's context. These cookies provide full account access and are high-value targets for exfiltration.
[EXTERNAL_DOWNLOADS]: The skill relies on and recommends tools from community-maintained GitHub repositories rather than official vendors. Specifically, it references xpzouying/xiaohongshu-mcp and Nemo2011/bilibili-api, which are not verified or trusted organizations.
[COMMAND_EXECUTION]: The workflow involves executing shell commands for wechatsync, biliup, and xiaohongshu-mcp (via local daemon), which grants the skill significant control over the local environment for publishing operations.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from source files (article.md) and passes it to multiple specialized sub-agents (e.g., @zhihu-strategist).
Ingestion points: Reads content from article.md in Step 2 of the workflow.
Boundary markers: The skill lacks explicit instructions or markers to distinguish between legitimate content and potentially malicious embedded instructions in the source file.
Capability inventory: The skill has the ability to execute shell commands (wechatsync), perform network requests (curl to a local API), and upload files to remote servers.
Sanitization: There is no evidence of sanitization or filtering applied to the content before it is interpolated into prompts for the specialist agents or used in tool calls.

agency-multi-platform-publisher