skills/amit-t/skills/session-feedback/Gen Agent Trust Hub

session-feedback

Pass

Audited by Gen Agent Trust Hub on Oct 31, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: No malicious patterns were detected. The skill operates locally on conversation data and project memory directories. It does not perform network operations or access sensitive system files outside of its documented scope.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it ingests untrusted data from user turns and past feedback files. However, the design includes a robust mitigation through a mandatory confirmation step requiring the user to approve any remembered guidance before it can influence the agent's actions.
  • Ingestion points: Conversation history and feedback_*.md files in memory directories (e.g., ~/.claude/projects/).
  • Boundary markers: Structured markdown templates with specific headers and IDs are used to separate feedback items.
  • Capability inventory: File system access for reading, writing, and listing memory files via the agent's built-in toolset.
  • Sanitization: No technical sanitization or escaping is performed on user input, as the skill's purpose is to capture verbatim phrasing.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 31, 2026, 04:28 PM