object-writing

Pass

Audited by Gen Agent Trust Hub on Jul 21, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill parses user input via the $ARGUMENTS variable to route specific writing actions (e.g., metaphor, cliche). While this represents a surface for indirect prompt injection, the risk is negligible as the skill is limited to creative text generation within a local context. \n
  • Ingestion points: $ARGUMENTS variable in SKILL.md. \n
  • Boundary markers: Absent. \n
  • Capability inventory: Local file writing and template resolution. \n
  • Sanitization: Absent. \n- [REMOTE_CODE_EXECUTION]: The skill uses a dynamic path resolution mechanism to check for template overrides within the project directory (${CLAUDE_PROJECT_DIR}). This is a standard customization feature for agent skills and does not involve remote code execution or execution of untrusted external content. \n- [COMMAND_EXECUTION]: The skill identifies a mechanism for persisting artifacts to the file system based on paths defined in a local configuration (artifact-persistence.md). This is a documented functional requirement for the skill's purpose and does not exhibit malicious intent.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 21, 2026, 07:55 PM
Security Audit — agent-trust-hub — object-writing