skills/blunotech-dev/agents/slides/Gen Agent Trust Hub

slides

Warn

Audited by Gen Agent Trust Hub on Aug 12, 2026

Risk Level: MEDIUMPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill metadata lists the author as 'claudekit', which conflicts with the provided author context 'blunotech-dev'. This discrepancy in metadata is considered deceptive.
  • [PROMPT_INJECTION]: The skill ingests user input via the $ARGUMENTS variable to route subcommands and define slide tasks. This represents an attack surface for indirect prompt injection. Ingestion points: $ARGUMENTS in SKILL.md and references/create.md. Boundary markers: XML-style tags (<args>, <task>) are used. Capability inventory: Generates HTML/JS and references a local search script. Sanitization: No explicit sanitization or validation of input is performed.
  • [COMMAND_EXECUTION]: The documentation contains examples for executing a local Python script at .claude/skills/design-system/scripts/search-slides.py to search knowledge base files.
  • [EXTERNAL_DOWNLOADS]: The HTML template fetches the Chart.js library from the JSDelivr CDN, which is a well-known and trusted service for front-end libraries.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 12, 2026, 09:06 PM
Security Audit — agent-trust-hub — slides