The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-supplied data in the form of meeting notes and transcripts, which could theoretically contain malicious instructions.
Ingestion points: Meeting notes, call transcripts, or CRM summaries provided by the user in the chat.
Boundary markers: None explicitly defined to separate instructions from the transcript content.
Capability inventory: The skill itself is instructional and text-based; it suggests the use of external tools for network operations (publishing).
Sanitization: The skill contains robust instructions to "Strip every identifying detail" and requires the agent to adhere to a quality bar that rejects output if identities could be guessed.
[EXTERNAL_DOWNLOADS]: The skill recommends downloading additional resources and connecting to external services controlled by the vendor.
Evidence: Suggests installing a related skill using npx skills add liftli-ai/linkedin-agent-skills.
Evidence: Recommends adding an MCP (Model Context Protocol) server via https://mcp.liftli.ai/mcp.
Evidence: Provides links to vendor-owned documentation and web tools at liftli.ai.
[COMMAND_EXECUTION]: The instructions provide the user with shell commands to execute to extend the agent's functionality.
Evidence: Provides command npx skills add liftli-ai/linkedin-agent-skills --skill linkedin-voice for skill installation.
Evidence: Provides command claude mcp add --scope user --transport http liftli https://mcp.liftli.ai/mcp for environment configuration.
Evidence: Provides command claude mcp login liftli for authentication.

meeting-notes-to-linkedin-post