verse-craft

Pass

Audited by Gen Agent Trust Hub on Nov 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill employs the 'story' CLI tool for project management tasks, such as initializing projects ('story init'), adding chapters or epigraphs ('story add'), reindexing the project ('story reindex'), and performing word count updates ('story wordcount'). These commands operate locally on the project's files.
  • [DYNAMIC_EXECUTION]: Includes instructions for a fallback maintenance method using 'node' to execute a bundled script ('story.js'). The workflow specifies resolving the script path relative to the skill directory and includes explicit security warnings to avoid runtimes like Bun that might automatically execute code from local configuration files.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from project files (metadata, character voice notes, and chapter content) to guide the generation and critique of verse, creating a theoretical attack surface.
  • Ingestion points: Accesses 'story.md', character definition files, and existing chapter text files during the drafting and critique phases.
  • Boundary markers: The skill does not mandate the use of delimiters or explicit instructions for the agent to ignore potentially malicious commands embedded in the ingested story content.
  • Capability inventory: The skill possesses the capability to write to the filesystem (via word count updates) and execute shell commands (via the 'story' CLI).
  • Sanitization: Content retrieved from project files is not subject to explicit sanitization, escaping, or validation before being included in the agent's context.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 14, 2026, 05:48 PM