The Agent Skills Directory

[SAFE]: The skill uses codebase exploration tools (Glob, Grep, Read, LSP) to gather context for providing architectural feedback. This is standard behavior for a development-focused agent and does not involve unauthorized file access or credential exposure.
[SAFE]: Web search and fetching operations are used to research library documentation and performance characteristics. These activities are limited to information retrieval from well-known sources and do not include the execution of remote scripts or unverified packages.
[SAFE]: The instructions provide behavioral guidance for a "relentless" role-play persona. These prompts do not contain patterns associated with prompt injection, such as instructions to ignore safety filters or bypass system constraints.
[SAFE]: While the skill interacts with untrusted data (external web pages and user-provided code), this indirect prompt injection surface is managed through the agent's standard conversational interface without granting the skill autonomous code execution or sensitive data transmission capabilities.

grill-me