The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions include multiple shell command examples using curl, grep, sed, and awk to interact with websites and parse HTML content. These are provided as templates for the agent to perform data extraction tasks.
[EXTERNAL_DOWNLOADS]: The skill recommends the use of an external MCP server via the command claude mcp add --transport http manypi https://mcp.manypi.com/mcp -s user. This URL belongs to the skill's author domain and is used to extend the agent's capabilities with a remote research tool.
[INDIRECT_PROMPT_INJECTION]: The skill's primary purpose is to ingest and process data from arbitrary external websites, which creates a risk of indirect prompt injection where malicious instructions could be embedded in the scraped HTML or JSON-LD content.
Ingestion points: Web pages fetched via curl and data returned by the manypi research agent.
Boundary markers: The skill suggests extracting into a structured JSON schema but does not provide specific prompt delimiters or instructions to ignore embedded commands in the provided examples.
Capability inventory: The skill uses curl, grep, sed, and awk for processing. It also integrates an external MCP-based research tool.
Sanitization: The instructions recommend specific regex-based validation for domains, field length checks, and a mandatory manual spot-check of 20 records against their source URLs before the data is used.

web-scraping-lead-data