perfectpixel

Fail

Audited by Gen Agent Trust Hub on Jun 30, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The scripts/install.sh script downloads pre-built binaries from the gykim80/perfectpixel-studio GitHub repository, which is not a trusted vendor.- [REMOTE_CODE_EXECUTION]: The skill facilitates the download and execution of a binary (ppgen) from an untrusted external source. It also supports cloning a repository and compiling code using go build if the binary download fails.- [COMMAND_EXECUTION]: The skill executes shell scripts (install.sh) and the ppgen binary with arguments derived from user-supplied descriptions, presenting a risk of command injection.- [DATA_EXFILTRATION]: The skill requires users to provide sensitive API keys for services like Gemini, OpenRouter, and fal.ai, which are then passed as command-line arguments to the externally downloaded binary.- [PROMPT_INJECTION]: The skill takes user input for sprite descriptions and interpolates it into CLI flags for processing by AI models.
  • Ingestion points: SKILL.md (Step 2
  • -desc flag).
  • Boundary markers: Absent.
  • Capability inventory: Subprocess execution of external binary and shell commands.
  • Sanitization: Absent.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Jun 30, 2026, 04:49 AM
Security Audit — agent-trust-hub — perfectpixel