The Agent Skills Directory

[Data Exposure & Exfiltration] (SAFE): The brand_audit.py script performs local file reads to analyze color and font usage. It does not contain any network-facing code (no requests, urllib, or socket operations) and respects common directory exclusions like .git and node_modules.
[Unverifiable Dependencies & Remote Code Execution] (SAFE): All scripts use standard Python libraries (os, json, re, argparse). There are no external package requirements or remote script downloads.
[Dynamic Execution] (SAFE): No use of eval(), exec(), or runtime compilation was detected. The scripts use static regex patterns for analysis.
[Indirect Prompt Injection] (LOW): The brand_audit.py script reads external file content from a user-specified directory. While it includes file content snippets in its audit report, the risk is minimal as the script's output is purely informational and lacks privileged capabilities.
Ingestion points: Reads content from .html, .css, .md, and .txt files in scripts/brand_audit.py.
Boundary markers: None; the script outputs raw findings from files.
Capability inventory: Local file reading and stdout reporting; no subprocess or network calls.
Sanitization: None; however, the script only extracts specific hex codes and terminology matches.

branding