The Agent Skills Directory

[OBFUSCATION]: The skill's primary logic, instructions, and reference documents (e.g., references/article-state-contract.md.enc, scripts/validate_skill.py.enc) are provided in an encrypted format. This prevents any verification of the code's behavior or safety through static analysis.
[EXTERNAL_DOWNLOADS]: The skill instructions direct the user to use npx and uvx to install tools from the developer's ecosystem, specifically lovstudio and lovstudio-skill-helper.
[REMOTE_CODE_EXECUTION]: The skill utilizes uvx lovstudio-skill-helper decrypt to process encrypted files at runtime. This creates a dependency on remote code that dynamically handles the skill's execution environment.
[COMMAND_EXECUTION]: The SKILL.md file contains shell commands for the installation and decryption of the skill, which are intended to be executed by the user or the agent.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to operate on WeChat articles, which are external data sources that could contain malicious instructions.
Ingestion points: Reads and edits WeChat article content as described in SKILL.md.
Boundary markers: Verification of boundary markers or 'ignore' instructions is impossible because the reference files (references/*.md.enc) are encrypted.
Capability inventory: The skill manifest lists Python scripts (scripts/*.py.enc), but their specific file-system or network capabilities are obscured by encryption.
Sanitization: It is impossible to verify if the skill sanitizes article content before processing due to the encrypted nature of the validation scripts.

lov-wechat-article-operator