debug-mode
Pass
Audited by Gen Agent Trust Hub on Jun 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute local utility scripts including start-server.sh, stop-server.sh, and clear-log.sh. It also launches a background Node.js process (debug-server.js) to host a local HTTP service on 127.0.0.1:3847 for log aggregation.\n- [PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection because it instructs the agent to read and analyze runtime logs produced by the application being debugged. If the application logs untrusted data containing adversarial instructions, the agent might interpret them as valid commands.\n
- Ingestion points: Data is ingested from .claude/debug.log using tools like Read, Grep, and tail during the diagnostic phase.\n
- Boundary markers: The skill expects NDJSON formatting and uses [DEBUG H*] tags, but these do not constitute a secure boundary to prevent the execution of instructions found within log messages.\n
- Capability inventory: The agent has access to powerful tools including Bash, Write, and Edit, which increases the potential impact of a successful injection.\n
- Sanitization: There is no indication of sanitization or filtering applied to log contents before they are analyzed by the agent.
Audit Metadata