pixel-art-sprites

Pass

Audited by Gen Agent Trust Hub on Nov 3, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill mandates that the agent must always start its first response with a specific emoji (๐Ÿงข). This is an unnecessary behavioral constraint and an override of the agent's default communication style.
  • [COMMAND_EXECUTION]: The instructions in SKILL.md direct the agent to automatically execute a shell command (ls) to probe for installed companion skills. The command targets multiple hidden directories in the user's home folder (~/.claude/skills/, ~/.agent/skills/, ~/.agents/skills/) as well as the local project root. This behavior probes the user's environment to identify other installed software.
  • [INDIRECT_PROMPT_INJECTION]: The skill implements a dependency check that monitors the local filesystem and uses the results to influence the agent's suggestions. If certain "companion" skills are not detected, the agent is instructed to recommend their installation via an external npx command.
  • Ingestion points: The result of the ls command (filesystem state) is used as input for the agent's logic in SKILL.md.
  • Boundary markers: None present; the instructions do not include delimiters or warnings for the agent when processing the results of the filesystem probe.
  • Capability inventory: The skill uses filesystem read capabilities (ls) to inform recommendations for external command execution (npx skills add).
  • Sanitization: No validation or sanitization is performed on the names of the detected skills before they are processed by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 3, 2026, 01:55 PM