The Agent Skills Directory

[SAFE]: The skill consists exclusively of markdown instructions and formatting guidelines. No executable scripts, shell commands, or binary files were detected in any of the provided files.
[DATA_EXPOSURE]: No hardcoded credentials, API keys, or paths to sensitive configuration files (e.g., .ssh, .aws, .env) are present. The skill targets the generation of public-facing financial research.
[REMOTE_CODE_EXECUTION]: There are no patterns involving the download of external code or the execution of remote scripts. No package managers (npm, pip) or dynamic execution functions (eval, exec) are used.
[PROMPT_INJECTION]: The instructions are focused on adherence to professional writing styles and do not contain attempts to override safety filters, bypass constraints, or extract system prompts.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external PDF documents (yearly outlooks). While processing external data is an entry point for indirect instructions, the skill does not possess any high-privilege capabilities—such as network exfiltration or file system modification—that would allow for an exploit to cause harm.
Ingestion points: SKILL.md references uploaded PDF files (2024, 2025, 2026 versions) as data sources.
Boundary markers: None explicitly defined in the prompts.
Capability inventory: No subprocess calls, network ops, or file writes detected.
Sanitization: No specific sanitization or validation logic is applied to the content of the referenced PDFs.

commodity-research-outlook