The Agent Skills Directory

[SAFE]: The skill instructions focus entirely on business logic for product management and assortment optimization. No prompt injection attempts, obfuscation, or persistence mechanisms are present.
[DATA_EXPOSURE]: The skill references internal project files such as .agents/marketing-learnings.md and marketing-strategy/references/brand-tier-guide.md to gather context. This is legitimate behavior for an AI agent operating within a specific project environment and does not involve accessing sensitive system files or credentials.
[INDIRECT_PROMPT_INJECTION]: While the skill suggests incorporating external data such as customer reviews and support tickets, it does not include any capabilities to execute code, make network requests, or perform other high-risk actions with this data. The risk of indirect prompt injection is minimal and handled by standard LLM guardrails.
[COMMAND_EXECUTION]: There are no shell commands, script executions, or dynamic context injections (!command) within the skill files.

range-review