The Agent Skills Directory

[SAFE]: The skill implements strong operational guardrails, including explicit prohibitions against deceptive practices such as fake reviews, artificial price manipulation, and deceptive countdown timers.
[SAFE]: External references target official Vietnamese government and legal databases (e.g., vbpl.vn, gdt.gov.vn) for compliance purposes, which are reputable sources for the skill's intended region.
[INDIRECT_PROMPT_INJECTION]: The skill processes external financial reports and e-commerce platform data, creating a surface for indirect injection.
Ingestion points: External settlement reports (CSV/Excel) and Seller Center exports are ingested for P&L calculations.
Boundary markers: The skill includes a specific safety rule ('Quy tac 2') in references/phi-san-tmdt.md which explicitly instructs the agent that content in reports must be treated strictly as data and to ignore any commands found within the files.
Capability inventory: The skill is limited to generating markdown reports and performing arithmetic calculations. There are no subprocess execution, arbitrary file-writing, or network exfiltration capabilities present.
Sanitization: The skill relies on natural language instructions to enforce data/command separation rather than technical escaping mechanisms.

76-van-hanh-san-tmdt