linkedin-humanizer
Pass
Audited by Gen Agent Trust Hub on Oct 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill uses local Python scripts (tell_audit.py, emoji_audit.py, voice_fingerprint.py) that rely exclusively on standard libraries such as argparse, json, re, and statistics. No external dependencies or remote script executions are present.
- [SAFE]: No network operations or data exfiltration patterns were detected. The scripts read local text files provided by the user and output analysis reports to stdout.
- [SAFE]: The skill includes defensive patterns to identify and remove AI-assistant residue (e.g., citation tokens, preambles, and placeholders), which helps prevent accidental publishing of raw model outputs.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user drafts via the --input argument in tell_audit.py. Ingestion points: The scripts read user drafts from the file system or stdin. Boundary markers: Instructions in SKILL.md guide the agent to remain truthful, use user-provided facts, and avoid inventing details. Capability inventory: Limited to regex analysis and agent-led rewriting based on audit results. Sanitization: No explicit code-based sanitization for the draft content beyond matching against specific audit patterns.
Audit Metadata