story-maintenance
Pass
Audited by Gen Agent Trust Hub on Nov 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONOBFUSCATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from manuscripts and Word documents via the import and notes commands (Ingestion points: SKILL.md, references/editing-commands.md). Capabilities include executing shell commands via the bundled CLI and writing to the local filesystem. While the skill instructs the agent to treat external text as non-authoritative (Sanitization), the processing of arbitrary natural language from these sources remains an attack surface.
- [COMMAND_EXECUTION]: The skill relies on executing a local CLI utility (scripts/story.js) via the Node.js runtime to perform structure validation, reindexing, and other project maintenance tasks.
- [OBFUSCATION]: Automated static analysis flagged potential steganographic Unicode characters within the execution scripts. Although documentation in references/conventions.md suggests the skill is designed to detect and report invisible characters in user-provided text as a security feature, their presence within the skill's own logic files is a potential obfuscation vector.
Audit Metadata