skills/zzci/skills/spectrum-audit/Gen Agent Trust Hub

spectrum-audit

Warn

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: MEDIUMMETADATA_POISONINGEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [METADATA_POISONING]: The skill metadata identifies the author as 'Adobe' and provides a link to an official Adobe website, which contradicts the platform-provided author identity 'zzci'. This discrepancy is misleading and may cause users to incorrectly assume the skill is an official Adobe product.
  • [EXTERNAL_DOWNLOADS]: The skill recommends installing a supplementary skill from the official Adobe domain ('https://react-spectrum.adobe.com'). Although Adobe is a well-known service, the recommendation comes from a skill with questionable author attribution.
  • [COMMAND_EXECUTION]: The skill directs the agent to execute 'npm view @react-spectrum/s2 version' to check for updates from the official npm registry.
  • [INDIRECT_PROMPT_INJECTION]: The skill's primary function involves processing untrusted codebase content, which creates a surface for indirect prompt injection.
  • Ingestion points: The skill reads project source files and configuration files (e.g., 'src/**/*.{tsx,jsx,ts,js}', 'package.json', lockfiles) as part of its audit process.
  • Boundary markers: No explicit delimiters or boundary instructions are used to separate the auditing instructions from the code being analyzed.
  • Capability inventory: The skill reads local files and writes a markdown report ('SPECTRUM-AUDIT.md'); it recommends external command execution for library versioning and skill installation.
  • Sanitization: The skill relies on pattern matching and grepping to identify violations and does not explicitly sanitize the analyzed source content.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 3, 2026, 08:11 PM
Security Audit — agent-trust-hub — spectrum-audit