skills/aws-samples/sample-improve-agent-token-efficiency-skill/improve-agent-token-efficiency/Gen Agent Trust Hub
improve-agent-token-efficiency
Pass
Audited by Gen Agent Trust Hub on Nov 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and analyze untrusted third-party codebases, which can contain malicious prompts or logic.
- Ingestion points: The skill reads files from user-provided code folders or URLs during the 'Understand existing agent' and 'Discover Improvement Opportunities' phases.
- Boundary markers: There are no explicit instructions or delimiters to ensure the model treats instructions found within the reviewed code solely as data.
- Capability inventory: The skill possesses significant capabilities, including generating and running a benchmarking script (run_agent_metrics.py) and modifying the agent's source code in Phase 5.
- Sanitization: The instructions do not specify any sanitization or validation of the untrusted code content before processing.
- [COMMAND_EXECUTION]: The skill facilitates the generation and local execution of a benchmarking script.
- Evidence: Phase 1b (01b_capture_baseline.md) and Phase 6 (06_validation.md) instruct the agent to create and run run_agent_metrics.py to exercise the target agent and record usage metrics.
- [DYNAMIC_EXECUTION]: The skill dynamically generates the logic for the metrics runner based on the specific framework of the agent being analyzed.
- Evidence: The skill provides a skeleton for run_agent_metrics.py and instructs the agent to adapt the entry points and usage-reading logic at runtime to match the target code.
Audit Metadata