dev-swarm-screen-snapshot

Pass

Audited by Gen Agent Trust Hub on Mar 13, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through the processing of screen capture data.
  • Ingestion points: The agent retrieves image snapshots from http://127.0.0.1:9090/snapshot.png as specified in SKILL.md.
  • Boundary markers: Absent. There are no instructions to disregard or isolate text found within the captured images.
  • Capability inventory: The skill has the capability to execute shell commands using uv run in SKILL.md.
  • Sanitization: Absent. The skill does not specify any sanitization or validation of the text content extracted from the visual snapshots.
  • [COMMAND_EXECUTION]: The skill executes a local Python script to start a background screen streaming process.
  • Evidence: The command uv run --directory {PROJECT-ROOT}/dev-swarm/py_scripts screen_stream.py --monitor -1 --background --fps 30 --host 127.0.0.1 --port 9090 is triggered in SKILL.md.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 13, 2026, 08:28 PM