The Agent Skills Directory

[SAFE]: The skill provides structured guidance for academic tasks and does not contain any prompt injection attempts designed to bypass safety filters or reveal system prompts.- [SAFE]: The included Python scripts perform localized analysis on manuscripts and state files. These scripts do not initiate network connections, access sensitive system files (e.g., SSH keys), or execute arbitrary code from the internet.- [SAFE]: External URLs in the documentation refer to well-known academic publishers (AGU, Wiley) and government agencies (FEMA) for reference purposes.- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided manuscripts and reviewer comments, which represents an indirect injection surface. However, the risk is mitigated by the skill's reliance on 'writing contracts' and deterministic audit scripts that enforce specific logic over free-form generation. (Ingestion points: manuscripts in .docx/.md/.txt formats, manuscript state JSONs; Boundary markers: Use of 'managed-project mode' and 'writing contracts' to limit processing scope; Capability inventory: Local file read/write and script execution for text auditing; Sanitization: Uses standard libraries for JSON and XML parsing with no unsafe evaluation of user data). Severity: SAFE.

academic-writing-skills-zh-tw