The Agent Skills Directory

[SAFE]: The skill consists entirely of natural language instructions in Markdown format and does not include any scripts, commands, or executable logic.
[DATA_EXFILTRATION]: No network requests, external URLs, or credential access patterns were found. The skill operates solely on the text context provided to the agent.
[PROMPT_INJECTION]: The instructions do not contain markers intended to bypass safety filters or override system-level constraints. The focus remains strictly on writing style (e.g., '棱角模式', '锐利模式').
[OBFUSCATION]: No hidden content, encoded strings (Base64/Hex), or deceptive character substitutions (homoglyphs/zero-width characters) were detected.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided text for rewriting.
Ingestion points: User input processed according to the '棱角', '锐利', or '密度' modes defined in SKILL.md.
Boundary markers: Absent.
Capability inventory: None; the skill does not use tools, perform shell commands, or write files.
Sanitization: Absent.
Conclusion: Because the skill lacks any capabilities to execute code or access external systems, the surface for indirect prompt injection is benign.

style-blade