fixing-claude-export-conversations
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted .txt files exported from Claude Code, which serves as a data ingestion surface for the agent. * Ingestion points: The scripts/fix-claude-export.py script takes a user-supplied input file. * Boundary markers: The script identifies message blocks using prefixes like ❯ and ● but does not implement delimiters or ignore-instructions to isolate content from potential instructions. * Capability inventory: The skill has the ability to read and write files using pathlib and uses the uv run command for script execution. * Sanitization: The script performs regex-based formatting and structural cleanup but does not sanitize the natural language content for potential embedded instructions.
Audit Metadata