The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes lead data from external and untrusted sources including Google Sheets, CSV files, and LinkedIn profile URLs. This ingestion represents a potential attack surface where malicious instructions could be embedded in data fields (e.g., a LinkedIn bio or company description) to influence the agent's behavior.
Ingestion points: The skill reads data from user-provided CSV paths, Google Sheet URLs, and LinkedIn URLs.
Boundary markers: The instructions do not define specific delimiters or "ignore embedded instructions" warnings when passing lead data to the parallel qualification subagents.
Capability inventory: The agent has the ability to execute shell commands (running the enrichment script), perform network operations (Apify API, web search), and write to external services (Google Sheets via Rube).
Sanitization: No explicit sanitization or filtering of the lead data is described before it is interpolated into the prompts for subagents.
[EXTERNAL_DOWNLOADS]: The skill relies on external infrastructure and libraries for its enrichment phase.
The scripts/enrich_leads.py script requires the third-party requests library.
The skill uses the Apify platform, specifically the supreme_coder~linkedin-profile-scraper actor, to perform batch profile enrichment.
[COMMAND_EXECUTION]: The skill instructions direct the agent to execute a local Python script (python3 skills/lead-qualification/scripts/enrich_leads.py) using a user-provided file path as an argument. While standard for the workflow, this involves shell execution of a script with external input.
[DYNAMIC_EXECUTION]: During the intake phase, the skill generates a structured qualification prompt based on user conversation and saves it as a markdown file. This file is then dynamically used as the instruction set for subsequent qualification tasks, creating an instructions-generating-instructions pattern.

lead-qualification