Security audit
Security checks for vulnerabilities and agentic risk
The skill is mostly a legitimate curriculum-document generator, but it keeps all intermediate files and logs and includes under-scoped remote PDF downloading.
Review this skill before installing if you will process sensitive school, student, or institutional documents. It should ideally ask before retaining debug files, define cleanup rules, and restrict remote PDF downloads to verified official sources with size and type checks.
SKILL.md:258Unrestricted URL Retrieval Enables Server-Side Request Forgery
SKILL.md:759Unconditional Retention of Intermediate Files and Logs Risks Sensitive Data Exposure
The workflow first prohibits guessing or using defaults for required fields, then later permits placeholders or inferred structures when key information is missing. This contradiction weakens safety and integrity controls because the agent may proceed with fabricated or inferred content that appears authoritative, causing silent data quality failures and enabling prompt-induced bypass of required validation gates.
The trigger list includes broad, generic educational terms such as 课程目标, 教学评价, 课程结构, and 课程建设, which can overlap with many benign education-related conversations. Overbroad activation can cause the wrong skill to run unexpectedly, pulling in unnecessary workflow behavior and increasing exposure to unintended file access, processing steps, or network retrieval logic contained in the skill.
The skill instructs the agent to fetch remote PDF files from ministry links when the user has not supplied a teaching standard. That expands the trust boundary from user-provided local inputs to network-retrieved content, creating risks such as SSRF-like misuse if URLs or index contents are tampered with, ingestion of malicious or unexpected documents, and nondeterministic behavior unrelated to the core task of generating a curriculum standard from supplied materials.
The skill says the learning-scenario template type must be selected by the user and must not be auto-selected, but later fallback logic substitutes another template or simplified version automatically when template reading fails. That creates a policy bypass where a mandatory user decision is replaced by implicit system behavior, reducing predictability and potentially producing noncompliant output without informed consent.
No suspicious patterns detected.