The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains strict operational instructions to ensure the agent remains focused on root-cause analysis (e.g., 'STOP. Run problem-analysis first'). These are legitimate safety and quality constraints for the business logic and do not attempt to bypass core AI safety filters.
[DATA_EXFILTRATION]: No network operations or credential harvesting patterns were found. The skill operates on local metadata files (e.g., '.agents/product-context.md') which is standard for this agentic workflow.
[EXTERNAL_DOWNLOADS]: The skill references other skills from the same author (e.g., 'hungv47/marketing-skills') and well-known documentation. There are no downloads from untrusted third-party servers or remote script executions.
[COMMAND_EXECUTION]: The provided files consist of markdown instructions for agents and reference materials. There are no shell commands, subprocess calls, or privilege escalation attempts.
[OBFUSCATION]: Comprehensive scanning for Base64 encoding, zero-width characters, homoglyphs, and hidden text patterns yielded no results. All content is transparent and readable.

solution-design