The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The documentation suggests an alternative configuration using the npx command to download the @porkll/siyuan-mcp package from the npm registry. This introduces a dependency on code from an unverified third-party account that is not part of the vendor's official resources.\n- [REMOTE_CODE_EXECUTION]: The recommendation to execute npx -y @porkll/siyuan-mcp allows for the fetching and running of arbitrary code from a remote repository at runtime.\n- [COMMAND_EXECUTION]: The skill provides instructions for using a SQL endpoint (/api/query/sql). Although it advises restricting use to SELECT statements, the endpoint itself may support destructive commands (such as DELETE or DROP) depending on the server configuration, creating a risk of unauthorized data modification or loss.\n- [PROMPT_INJECTION]: The skill retrieves and processes user-generated note content, which creates a surface for Indirect Prompt Injection. \n
Ingestion points: Retrieved note content and search results from /api/block/getBlockKramdown, /api/export/exportMdContent, and /api/search/fullTextSearchBlock.\n
Boundary markers: The instructions do not define delimiters or specific warnings to ignore instructions embedded within the retrieved note data.\n
Capability inventory: The skill utilizes curl to perform file and block management tasks including creation, modification, and deletion as documented in SKILL.md.\n
Sanitization: There is no mention of content validation or sanitization for the data retrieved from the API before it is consumed by the agent.

siyuan