creative-plan

Warn

Audited by Gen Agent Trust Hub on Oct 27, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill performs dynamic path computation for file system access at runtime.
  • It calculates PLUGIN_ROOT by performing string manipulation on the Base directory environment variable.
  • It constructs DOMAIN_DIR using a domain variable extracted from an untrusted project file (.creative/config.yaml).
  • The agent then executes Read operations on files within these computed paths (e.g., $DOMAIN_DIR/domain.yaml, $DOMAIN_DIR/plan-guide.md). This logic lacks validation to ensure the resulting paths remain within intended directory boundaries.
  • [INDIRECT_PROMPT_INJECTION]: The skill's primary planning logic and quality control standards are ingested from external, potentially user-controlled files.
  • Ingestion points: The skill reads domain.yaml and plan-guide.md from the project directory to define its planning layers, iteration logic, and output formats.
  • Boundary markers: There are no explicit delimiters or instructions to ignore embedded commands when the agent processes the contents of these guide files.
  • Capability inventory: The skill utilizes Read, Write, Edit, and AskUserQuestion tools, allowing the ingested instructions to potentially influence file system modifications and user interactions.
  • Sanitization: The instructions do not specify any validation, escaping, or filtering for the content loaded from the domain configuration or planning guides.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 27, 2026, 11:39 AM