The Agent Skills Directory

[SAFE]: The skill is designed for academic prose generation and stylistic adjustment. No malicious code, exfiltration, or persistence patterns were detected.- [COMMAND_EXECUTION]: The skill includes instructions to run a local script, scripts/humanize_check.py, which validates the completeness of a humanization matrix against project manuscript files. The script performs standard file reading and writing within the local workspace and does not exhibit dangerous behaviors such as privilege escalation or network communication.- [PROMPT_INJECTION]: The skill defines tiered constraints in references/humanize-tiers-en.md and references/humanize-tiers-zh.md to guide the AI's writing style. These instructions are focused on sentence structure, paragraph templates, and connector frequency, and do not contain attempts to bypass safety filters or override core agent behavior.- [DATA_EXFILTRATION]: Analysis of the Python validation script and the instruction files shows no evidence of network activity, credential harvesting, or sensitive file access (e.g., SSH keys or environment variables).

paper-spine-humanize