The Agent Skills Directory

[PROMPT_INJECTION]: No evidence of prompt injection attempts was found. The instructions use standard natural language for task definition and do not attempt to override safety filters or disregard prior constraints.
[DATA_EXPOSURE]: No hardcoded credentials, sensitive file paths, or data exfiltration patterns were detected. The skill primarily defines templates for business planning documentation.
[EXTERNAL_DOWNLOADS]: The skill mentions external services like Hugging Face (e.g., "Hugging Face estimator" in saas-ai-sustainability-and-ethics/SKILL.md) and standard frameworks like the UN SDGs and IFC Performance Standards. These are well-known, trusted resources used for reference and do not involve untrusted remote code execution.
[COMMAND_EXECUTION]: No shell commands, subprocess calls, or system-level operations are present in any of the analyzed files.
[OBFUSCATION]: No obfuscation techniques such as Base64-encoded instructions, zero-width characters, or homoglyph attacks were identified.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted inputs such as business models, AI architecture details, and sector contexts (e.g., in SKILL.md and saas-ai-sustainability-and-ethics/SKILL.md). However, the skill lacks any dangerous capabilities—such as file writing, network transmission, or code execution—that could be exploited via indirect injection. The analysis concludes that this surface is not exploitable.
[DYNAMIC_EXECUTION]: No dynamic code execution patterns like eval(), exec(), or runtime compilation were found.
[DYNAMIC_CONTEXT_INJECTION]: No use of the dynamic command injection syntax (!command) was detected.

16-sustainability-strategy