The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): The skill defines a large ingestion surface with high-privilege side effects.
Ingestion points: SKILL.md specifies processing of data from web scraping, PDF extraction, and API integrations.
Boundary markers: There are no specified delimiters or instructions to treat external data as untrusted or to ignore embedded instructions.
Capability inventory: SKILL.md grants the agent code execution, API calls, database queries, and web search capabilities.
Sanitization: Although it mentions guardrails and PII detection, these are descriptive capabilities rather than enforced sanitization routines for input data.
[Command Execution] (HIGH): The skill explicitly claims the capability for code execution and tool integration in SKILL.md, providing a powerful primitive that can be abused if the agent is manipulated via malicious data.

ai-engineer