story-short-write
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill executes multiple local scripts (
scripts/check-ai-patterns.js,scripts/check-degeneration.js,scripts/author_memory_commit.py, etc.) to automate word counting, punctuation normalization, and style verification. These scripts are purely functional and operate only on local files within the user's workspace. - [SAFE]: Data persistence for user preferences ('author memory') is managed locally via JSON and Markdown files in the
.story/作者记忆/directory. The management script (author_memory_commit.py) uses standard libraries and has no network capabilities. - [INDIRECT_PROMPT_INJECTION]: The skill has a defined surface for indirect prompt injection as it ingests external story analyses and user-provided drafts from the
拆文库/and workspace directories. However, the risk is minimized by the skill's highly specific workflow, which uses dedicated validation scripts to ensure the output remains focused on story generation rather than obeying embedded commands. All ingested data is treated as creative reference material. - [SAFE]: Instructions provided in
SKILL.mdand references focus on creative writing techniques (e.g., 'hooks', 'emotional pull', 'reversal patterns') and do not attempt to override AI safety filters or extract system prompts.
Audit Metadata