The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill relies on extensive web research, fetching full articles from external sources to validate business ideas. This creates a surface where malicious instructions hidden in web pages could potentially influence the agent's subsequent analysis or actions.
Ingestion points: External data enters the agent context through web_search and web_fetch calls described in SKILL.md (Section 3).
Boundary markers: The instructions do not define clear delimiters (like XML tags) or safety warnings to distinguish untrusted web content from the skill's core instructions.
Capability inventory: The skill has access to web search, web fetching, and shell execution for its bundled Python script.
Sanitization: There is no mention of filtering or sanitizing the content retrieved from external URLs before processing.
[COMMAND_EXECUTION]: The skill utilizes a bundled Python script (scripts/market_analyzer.py) to calculate market metrics. The agent is instructed to write a JSON file and execute this script via the shell. While the script itself is local to the skill, it operates on data derived from potentially untrusted external web sources.

startup-validator