WriteStory

Warn

Audited by Gen Agent Trust Hub on Mar 5, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill contains mandatory instructions to execute a shell command (curl -s -X POST http://localhost:8888/notify...) immediately upon invocation. This automated background process is used for 'Voice Notifications' but represents arbitrary command execution.
  • [COMMAND_EXECUTION]: The skill performs filesystem operations within the user's home directory (~/.opencode/skills/ and ~/.opencode/plans/). It reads configuration files and writes story plans (Story Bibles) to these paths, which grants the skill persistent access to user data areas.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its core workflow design.
  • Ingestion points: The Interview.md workflow specifically 'Consumes Available Input' from the user, including notes, outlines, and character descriptions.
  • Boundary markers: There are no defined delimiters or instructions to ignore embedded commands within the processed story data.
  • Capability inventory: The skill has the ability to execute shell commands (curl), write to the filesystem, and spawn additional agents with dynamically generated prompts in the Explore and WriteChapter workflows.
  • Sanitization: The skill does not perform any validation or sanitization of user-provided content before using it to structure the narrative architecture or influence subsequent agent tasks.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 5, 2026, 07:38 AM