cubeloop-trace

Pass

Audited by Gen Agent Trust Hub on Sep 10, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute the cubeloop trace command (via uv run) to interact with local JSONL trace files, allowing for listing, viewing, and statistical analysis of agent runs.
  • [INDIRECT_PROMPT_INJECTION]: The skill provides a surface for indirect prompt injection because it reads and displays content from previous agent runs, which may contain malicious instructions embedded in user inputs or tool outputs.
  • Ingestion points: JSONL trace files containing full span trees, including prompts, tool inputs, and outputs (as described in SKILL.md under the view and follow commands).
  • Boundary markers: The skill does not specify the use of delimiters or specific safety instructions to prevent the agent from interpreting instructions found within the logs.
  • Capability inventory: The skill uses CLI subcommands (ls, view, follow, stats, convert) that ingest and process log data, which is then presented to the agent context.
  • Sanitization: No sanitization or filtering of the trace data is described in the skill instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 10, 2026, 11:09 AM
Security Audit — agent-trust-hub — cubeloop-trace