coding-agent-sessions
Warn
Audited by Snyk on Jul 27, 2026
Risk Level: MEDIUM
Full Analysis
MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).
- Third-party content exposure detected (high risk: 0.78). Required workflow scans local outsider-authored transcript files (e.g., user prompts embedded in Claude/other agent JSONL under
~/.claude/.../projects/.../subagents/**/agent-*.jsonl), parses their free-text content (task/prompt) and emits it into the tool output that the agent would incorporate into its LLM context, creating an indirect prompt-injection channel.
Issues (1)
W011
MEDIUMThird-party content exposure detected (indirect prompt injection risk).
Audit Metadata