stylekit-style-prompts

Pass

Audited by Gen Agent Trust Hub on Feb 25, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill relies on local script orchestration through the use of subprocesses. Scripts such as run_pipeline.py, benchmark_pipeline.py, and ci_regression_gate.sh execute other local Python and shell scripts to automate the prompt engineering workflow.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8). It ingests untrusted user queries and structured reference data (JSON or files) which are then interpolated into complex prompts for downstream coding assistants.
  • Ingestion points: scripts/generate_brief.py (via --query, --reference-file, and --reference-json arguments).
  • Boundary markers: The generated prompts use descriptive headers but lack strict delimiters to isolate user-provided data from system instructions.
  • Capability inventory: The skill can execute local Python scripts and shell commands via subprocess.run as part of its pipeline.
  • Sanitization: The scripts perform tokenization and regex-based filtering for style matching but do not specifically sanitize inputs for prompt injection attack patterns.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 25, 2026, 11:09 AM