how-newsletters

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTIONNO_CODE
Full Analysis
  • [DATA_EXFILTRATION]: The skill contains a hardcoded absolute file path to the author's home directory.
  • Evidence: /Users/charlesdeist/Desktop/Documents/How <Newsletters>/ in SKILL.md.
  • Risk: This exposure leaks the author's system username ('charlesdeist') and directory hierarchy. It also makes the skill non-portable as it expects a specific local environment to function.
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by processing external data as behavioral instructions.
  • Ingestion points: The agent is directed to read case study files from the local /Users/charlesdeist/Desktop/Documents/How <Newsletters>/ directory to determine its writing voice.
  • Boundary markers: Absent. The instructions do not provide delimiters or warnings to ignore potential instructions embedded within the case study content.
  • Capability inventory: The agent uses file-reading capabilities to ingest content that directly influences its response generation and persona.
  • Sanitization: Absent. There is no validation or filtering mentioned to prevent the agent from executing instructions found within the case study library.
  • [NO_CODE]: The skill consists exclusively of Markdown documentation and instructions without any accompanying executable scripts or configuration files.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 12:38 AM