The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-supplied business cases and scenarios for strategic analysis. However, it does not employ explicit boundary markers (such as XML tags, triple backticks with warnings, or specific delimiters) to separate this untrusted user data from the agent's core simulation instructions.
Ingestion points: User-provided requests for business evaluation, as described in references/detailed-guide.md (e.g., 'avaliar uma plataforma de educacao').
Boundary markers: Absent from the instruction structure in both SKILL.md and references/detailed-guide.md.
Capability inventory: The skill primarily performs text analysis and generation using a provided framework; it does not explicitly invoke dangerous subprocesses, network operations, or file-writing tools in the provided files.
Sanitization: There is no evidence of input filtering, escaping, or validation to prevent the agent from following malicious instructions embedded within the user's business scenario data.

bill-gates