The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from user-provided drafts and local voice configuration files without explicit sanitization or boundary markers.
Ingestion points: Processes text drafts supplied in the chat session and reads voice configuration from .liftli/voice.md and ~/.liftli/voice.md.
Boundary markers: Absent. The instructions do not specify delimiters or warnings to ignore embedded commands within the analyzed drafts or voice files.
Capability inventory: The skill is designed to interact with tools like strategist_briefing and the linkedin-publish skill, creating a path where malicious text in a draft could potentially influence tool arguments or agent behavior.
Sanitization: Absent. The skill analyzes and interpolates the content directly without escaping or filtering.
[EXTERNAL_DOWNLOADS]: The instructions direct the user or agent to fetch and install additional skills and MCP servers from the vendor's remote infrastructure.
Evidence: Mentions npx skills add liftli-ai/linkedin-agent-skills and claude mcp add --scope user --transport http liftli https://mcp.liftli.ai/mcp.
[COMMAND_EXECUTION]: The skill provides specific shell commands for environment configuration and tool installation.
Evidence: Shell commands for adding MCP servers (claude mcp add ...) and installing skills (npx skills add ...) are embedded in the instructions for the agent to present to the user or execute.

ai-sounding-post-checker