inventory-allocation

Pass

Audited by Gen Agent Trust Hub on Mar 23, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by systematically scanning the project codebase for inventory-related logic and configurations. This allows instructions hidden in analyzed files to potentially influence the agent's behavior. Ingestion points: The skill reads various project files, including requirements, schemas, and configurations, across all phases of its analysis (Phase 1). Boundary markers: No delimiters or instructions are provided to the agent to ignore embedded instructions in the source data. Capability inventory: As a CLAUDE_CODE skill, the agent has access to file system operations and shell command execution. Sanitization: The skill does not implement validation or escaping of the content read from the codebase before processing.
  • [COMMAND_EXECUTION]: The 'SELF-EVOLUTION TELEMETRY' section instructs the agent to perform file system discovery and write operations within the ~/.claude/projects/ directory to log execution metadata. This involves automated searching and writing to platform-specific hidden folders.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 23, 2026, 10:57 AM
Security Audit — agent-trust-hub — inventory-allocation