linkedin-content-generator

Pass

Audited by Gen Agent Trust Hub on Jul 10, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill implements a personalization mechanism where user feedback is saved to 'scripts/memory.md' and later included in the prompt context. This creates a surface for indirect prompt injection where malicious feedback could influence the agent's future behavior.
  • Ingestion points: User feedback is captured via the '/feedback' command and stored in 'scripts/memory.md' by 'scripts/memory_manager.py' using verbatim CLI arguments.
  • Boundary markers: User-provided content is wrapped in XML-style '' tags in 'scripts/utils.py', but the content is not escaped and no specific instructions are provided to the agent to treat the memory as untrusted.
  • Capability inventory: The skill uses the 'Bash' tool to execute local Python scripts that generate prompts for the agent to follow.
  • Sanitization: There is no evidence of input validation, escaping of control characters, or sanitization before data is written to the memory file or used in prompt construction.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 10, 2026, 05:11 AM
Security Audit — agent-trust-hub — linkedin-content-generator