The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses instructional language to guide the AI in creative writing and review. It does not contain patterns attempting to bypass safety filters, extract system prompts, or override agent behavior in a malicious way.
[DATA_EXFILTRATION]: There is no evidence of network operations (e.g., curl, fetch) or access to sensitive local files (e.g., .ssh, .env). All operations are limited to processing provided fiction outlines and canon.
[REMOTE_CODE_EXECUTION]: The skill does not download external scripts or install third-party packages. It references local commands like /use-novel-style for stylistic application within the platform's expected workflow.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data in the form of project canon, outlines, and chapter cards (SKILL.md). While it does not implement explicit boundary markers or sanitization, its functional capabilities are limited to prose generation and review, which does not grant an attacker access to system-level commands or sensitive data.
[OBFUSCATION]: The instructions and configuration are written in clear, plain text. No Base64, hex encoding, or zero-width character obfuscation was detected.

novel-use