premise-workshop

Pass

Audited by Gen Agent Trust Hub on Oct 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes a command-line interface tool named story and a fallback Node.js script located in a sibling directory (../story-maintenance/scripts/story.js) to perform project management tasks like initialization and indexing.
  • [COMMAND_EXECUTION]: The instructions include high-quality security guidance for the agent to mitigate shell injection risks. It explicitly mandates the use of single quotes for argument wrapping and provides specific escaping rules for different shell environments (POSIX vs. PowerShell), while warning against the use of cmd.exe due to its unsafe handling of special characters.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided text (story sparks, titles, and loglines). While this creates an attack surface for indirect prompt injection, the skill mitigates this by instructing the agent to treat user text as data rather than instructions and providing the aforementioned command sanitization rules.
  • Ingestion points: User-provided story sparks, titles, and descriptions entered during the workshop workflow.
  • Boundary markers: The skill instructs the agent to wrap user values in single quotes during command execution.
  • Capability inventory: Subprocess execution via the story CLI and Node.js for project initialization, name checking, and indexing.
  • Sanitization: Explicit instructions on shell-specific character escaping and quoting are provided to the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 15, 2026, 03:07 PM