The Agent Skills Directory

[SAFE]: The skill is entirely composed of Markdown documentation and instructions. It does not contain any executable scripts, binary files, or configuration that triggers command execution.
[CREDENTIALS_UNSAFE]: The skill explicitly instructs the agent to detect and redact sensitive technical tokens such as API keys, passwords, and connection strings, replacing them with stable placeholders (e.g., <API_TOKEN_1>). This is a defensive security practice.
[EXTERNAL_DOWNLOADS]: No remote URLs, network fetch operations (curl/wget), or external package dependencies were detected. All references are local to the skill's directory (e.g., references/terminology-guide.md).
[DATA_EXFILTRATION]: The skill contains a specific privacy constraint stating that the reviewer 'must not send the prompt to an external service merely to judge wording.'
[INDIRECT_PROMPT_INJECTION]: While the skill processes untrusted user prompts as its primary function, it does not possess tools or capabilities (like file writing or network access) that could be exploited via indirect injection. The instructions emphasize preserving the user's original intent without reinterpretation, which acts as a guardrail against instruction overrides.
[OBFUSCATION]: No base64, hex encoding, zero-width characters, or homoglyph-based obfuscation techniques were found in any of the skill files.

english-prompt-review