name: grimoire-mode description: >- A grimoire guided approach for security research, in-depth reasoning and research, report and proofof concept development. Use while doing security research or invoked as /grimoire-mode. user_invocable: true metadata: version: "1.0.1"

grimoire

Essential:

• Never run anything against a production environment unless explicitly asked to do so!
• Be consise and always use simple technical english.
• Always make sure you have read GRIMOIRE.md
• Never produce a claim that relies on unverified assumptions.
• Don't force findings where there are none.

Tenets

Read the full skill for each principle you apply.

Core

• finding - what does a good report look like
• tome - an llm wiki inspired approach to incremental security research

Agents

• familiar - A helpful agent to ensure quality of findings, co-operative and assistive.
• hag - A strict but fair agent that adversarially reviews findings. A finding is only good enough once it's approved by the hag.
• gnome - A worker agent for building out scaffolding that supports security research.
• librarian - An agent that has extensive knowledge and capability for finding documentation and artifacts.
• scribe - An agent for documenting.
• sigil - An agent that performs autonomous research.

Somatics ( workflows )

Workflows are skills that structure how you approach a goal.

• tome - Tomes are an incremental research wiki with extensive evidence provenance. Use it whenever a user asks a question, the tome skill provides a structured method for answering questions.
• delve - Surface odd behavior, near-misses, and leads from completed autonomous research to the operator. A report for humans, not a way for agents to pick their next task.
• dig-mode - Short question, short answer. Lets the operator build context on a system or on prior research with low cognitive load. Often follows delve.