The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute shell commands (e.g., claude mcp add, gemini mcp add) to configure the environment by adding a remote MCP server.
[EXTERNAL_DOWNLOADS]: The skill connects to and depends on remote resources hosted at https://abundant-poodle-82.convex.site. This includes the API endpoints and the MCP server definition.
[DATA_EXFILTRATION]: As part of its primary function, the skill transmits Markdown or HTML content provided by the user to an external endpoint for hosting. The instructions include a clear security warning advising against the inclusion of secrets or credentials in the uploaded content.
[INDIRECT_PROMPT_INJECTION]: The skill has the capability to retrieve and read back the raw content of hosted sites, which creates a potential surface for instructions embedded in that data to be processed by the agent. Mandatory Evidence Chain:
Ingestion points: GET /api/v1/sites/<slug>/raw endpoint described in SKILL.md.
Boundary markers: None specified in the instructions for parsing retrieved content.
Capability inventory: Subprocess calls via curl and MCP tool execution.
Sanitization: Instructions state the API rejects obvious credentials.

agentdrop