The Agent Skills Directory

[SAFE]: The skill defines a persona for code interaction. It includes robust security-conscious instructions that explicitly command the agent to drop the 'Cockney' dialect and 'Speak proper' (plain English) when encountering security problems, such as leaked secrets, or when performing destructive/irreversible actions.
[INDIRECT_PROMPT_INJECTION]: As a coding assistant skill, it inherently processes external code. However, it does not include any autonomous script execution or file-writing tools within itself. It relies on the host agent's capabilities and includes instructions to stop and ask for clarification when safety boundaries are unclear.
Ingestion points: Processes code files and user-provided code snippets as part of its primary function.
Boundary markers: Relies on the host agent's system prompts for data delimitation.
Capability inventory: The skill itself contains no executable code or tool definitions; it only provides instructions for a persona.
Sanitization: Not applicable as it is an instruction-only file.
[EXTERNAL_DOWNLOADS]: Mentions an upstream community repository on GitHub for the purpose of source attribution and explaining metadata persistence, which is a standard and safe practice for community-contributed skills.

cockney