skill-creator
Pass
Audited by Gen Agent Trust Hub on Jul 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands using Python's
subprocessmodule to invoke thepiCLI tool for testing and optimization tasks (scripts/run_eval.py,scripts/improve_description.py). It also utilizeslsofto manage local network ports for its evaluation viewer (eval-viewer/generate_review.py). - [EXTERNAL_DOWNLOADS]: The evaluation report viewer (
eval-viewer/viewer.html) loads the SheetJS library fromcdn.sheetjs.com. This is a well-known service for client-side spreadsheet processing. - [INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface by ingesting and processing external data from subagent transcripts and user-provided feedback files. These ingestion points (e.g.,
feedback.json,transcript.md) are used to drive the agent's grading and improvement logic. While the skill possesses capabilities like shell command execution, the instructions provide a structured workflow for handling this data.
Audit Metadata