Skip to content
Back to skills

Roguelike

ASecurity

This file is read by the kit whenever a project has `Claude/docs/.genre = roguelike`. Same learning mechanism as Survival.

  • 3 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 1, 2026
ai-agentspython

Security analysis

A100/100

Pro scans all 3 files and shows the line behind each finding

Scanned October 6, 2026

npx -y skills add mimmo-the-root/UEFN-dream-bot-team --skill roguelike --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Roguelike?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Roguelike
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/mimmo-the-root-roguelike/badge)](https://www.skillsdirectory.com/skills/mimmo-the-root-roguelike)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
genre_slug: roguelike
status: draft
variants_draft: [run-based-solo, run-based-coop, meta-progression]
last_updated: 2026-09-30
pattern_counts: proven=0 confirmed=0 hypothesis=0 contested=0
---

# Genre Skill: Roguelike

This file is read by the kit whenever a project has `Claude/docs/.genre = roguelike`. Same learning mechanism as Survival.

## How this skill learns (Skill Harness)

This skill starts EMPTY on purpose and is never pre-written from general knowledge. It learns from
the maps the owner actually builds, one map at a time — no need to wait for 3 maps:

- After a playtest (or on request) the `skill-reflector` agent QUEUES lessons. Nothing is learned
  until the owner approves it on the **Skills** page of the Agent Console.
- Every pattern carries a confidence level computed from the owner's own maps:
  **hypothesis** (1 map) → **confirmed** (2 maps) → **proven** (3+ maps, or 2 + a retention metric).
  A map that does the opposite turns it into **contested**; nothing is silently overwritten.
- How to use the list below: PROVEN = follow as a rule. CONFIRMED = follow by default, say why if you
  deviate. HYPOTHESIS = only a suggestion, tell the owner it is unproven. CONTESTED = show both options
  and ask. If the list is empty, there is simply nothing learned yet — do not invent patterns.
- Private vs shareable: map names, task IDs and notes live in `local/` (never exported, never
  committed). `pack/patterns.json` holds only generalized patterns + support counts and is the one
  part that can be shared or merged from the community. Do not edit the block below by hand
  (`python Claude/hooks/skills_lib.py ...` regenerates it).

## Starting a new project of this genre

Read `references/starter.md` (structure and questions only, no pre-written rules), then the learned patterns below.

## Known variants (prototype placeholders — rename or drop as soon as real maps show otherwise)

**run-based-solo** (attempt-based runs, one player), **run-based-coop** (shared run in co-op),
**meta-progression** (persistent unlocks between attempts). See `references/variants.md`.


## Dependencies (one-way, never the reverse)

This skill can read/cite the Device Library (second-brain) and the retention skill
(`fortnite-retention-gamedesign`). Those must NEVER contain logic specific to this genre.

<!-- PATTERNS:BEGIN -->
## Learned patterns (generated — do not edit by hand)

These come from maps the owner actually built and approved. Patterns the owner has not approved are not listed. Match the confidence to how you use each one.

_Nothing learned yet. Patterns appear here only after the owner approves a lesson from a real map._
<!-- PATTERNS:END -->

Files in this skill

  • SKILL.md2.5 KB
  • pack/patterns.json77 B
  • references/variants.md603 B

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.