agent-self-evaluation
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is entirely focused on providing a structured process for agent self-evaluation. It includes documentation, examples, and a local Python script for heuristic analysis of text.
- [COMMAND_EXECUTION]: The skill includes a Python script (
scripts/evaluate.py) and suggests the use of platform hooks to runechocommands. The Python script performs harmless regex-based keyword matching to score text for accuracy, completeness, and clarity. The commands used in hooks are benign reminders to the agent. - [INDIRECT_PROMPT_INJECTION]: While the skill involves processing agent outputs and user requests (which are ingestion points for potentially untrusted data), the processing is limited to calculating quality scores using local heuristics. There is no evidence of the processed data being used to drive dangerous operations, and the overall risk surface is negligible.
- [NO_CODE]: The provided Python script uses only the standard library (
argparse,re,sys,dataclasses,typing) and does not perform network operations, sensitive file access, or dynamic code execution.
Audit Metadata