reproduction-trace-instrumenter
Pass
Audited by Gen Agent Trust Hub on Mar 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The 'scripts/replay_generator.py' script dynamically constructs a Python simulation script ('replay.py') based on events recorded in a JSON trace file. This generated script is used to visualize the execution flow and does not execute the logic of the original application.
- [PRIVILEGE_ESCALATION]: The skill uses 'chmod' to set executable permissions (0o755) on the generated 'replay.py' file, facilitating local execution of the reproduction script.
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect injection surface through the ingestion of execution traces. 1. Ingestion points: Data is read from 'trace.json' by 'scripts/replay_generator.py'. 2. Boundary markers: There are no explicit delimiters or instructions to ignore content within the trace data. 3. Capability inventory: The skill is capable of writing new Python scripts and modifying their file permissions. 4. Sanitization: The skill uses the 'repr()' function to sanitize variable values before interpolating them into the generated replay script.
Audit Metadata