The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: Not detected. The skill executes a bundled local script (scripts/humanizer_scorer.py) to analyze text. This script only uses standard Python libraries (sys, re, json, math, collections) and does not invoke subprocesses or perform network operations.
[DATA_EXFILTRATION]: Not detected. The skill reads local content and context files (.claude/product-marketing-context.md) within the workspace to guide the writing process. There are no functions capable of transmitting this data to external servers.
[OBFUSCATION]: Not detected. All instructions, checklists, and the Python scoring script are provided in clear, readable text. No Base64, zero-width characters, or homoglyph attacks were found.
[INDIRECT_PROMPT_INJECTION]: Low risk. While the skill processes user-supplied content (draft.md) which could theoretically contain malicious instructions, the skill's capabilities are restricted to text analysis and rewriting. It lacks the 'write' or 'network' privileges required to perform harmful actions if triggered by an injection.
[COMMAND_EXECUTION]: Safe. The SKILL.md contains an example command to run the scoring script. This is a standard part of the skill's functionality and does not involve elevated privileges or dangerous arguments.

content-humanizer