The Agent Skills Directory

[PROMPT_INJECTION]: The skill's primary function is to ingest and analyze untrusted text provided by users (prompts, plans, arguments), creating an attack surface for indirect prompt injection.
Ingestion points: Processes user-provided prompts, system instructions, and strategy documents shared during the session (SKILL.md).
Boundary markers: The skill instructions do not specify the use of delimiters, XML tags, or "ignore" directives to separate analyzed content from the agent's core instructions.
Capability inventory: The skill leverages external tools such as mcp__coldiq__search_companies and mcp__coldiq__find_emails for data retrieval.
Sanitization: The instructions lack guidance on sanitizing or escaping user-provided inputs before they are processed by the model.
[EXTERNAL_DOWNLOADS]: The documentation includes instructions to fetch and execute external packages from a public registry.
Evidence: The skill text provides the command npx -y @coldiq/mcp@latest to download and run the ColdIQ MCP server from the NPM registry.
[COMMAND_EXECUTION]: The skill is designed to interact with external tools to validate claims and perform network-based lookups.
Evidence: Mentions the use of mcp__coldiq__find_people and mcp__coldiq__find_phone to verify assumptions within user strategies.

devils-advocate