skills/shinpr/rashomon/knowledge-base/Gen Agent Trust Hub

knowledge-base

Pass

Audited by Gen Agent Trust Hub on Jul 26, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill features an inherent attack surface by ingesting and storing user-provided feedback and comparison results to influence future agent prompts. This is a primary function of the skill and is confined to the local project environment.
  • Ingestion points: Untrusted data enters the system through user feedback and comparison results, which are stored in the file .claude/.rashomon/prompt-knowledge.yaml.
  • Boundary markers: The data is stored in a structured YAML schema, though no specific delimiters or 'ignore' instructions are used to wrap the retrieved patterns during interpolation into future prompts.
  • Capability inventory: The skill is limited to reading and writing the local .claude/.rashomon/prompt-knowledge.yaml file; it has no network access or subprocess execution capabilities.
  • Sanitization: No explicit sanitization or validation of the pattern content is performed before storage or retrieval.
  • [DATA_EXPOSURE]: The skill reads and writes to a specific local file (.claude/.rashomon/prompt-knowledge.yaml) within the project root. This file contains learned project patterns and does not involve accessing sensitive system files or credentials.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 26, 2026, 10:29 PM
Security Audit — agent-trust-hub — knowledge-base