po4yka/llm-wiki-skills / skills/llm-wiki-skill-compiler
llm-wiki-skill-compiler
VerifiedCompile procedural knowledge from an LLM-Wiki into installable Agent Skills. Use when wiki pages, runbooks, checklists, recurring workflows, troubleshooting notes, or query answers should become SKILL.md files with evidence links and validation.
LLM-Wiki Skill Compiler
Goal
Turn reusable procedural knowledge from the wiki into small, triggerable Agent Skills without moving domain knowledge into hidden prompts.
When to use
- A wiki page or folder captures a repeated step-by-step workflow, runbook, checklist, or troubleshooting routine worth reusing across sessions or agents.
- A saved query answer keeps getting re-asked and should become a triggerable skill instead of a synthesis note.
- An existing
skills/directory needs a new candidate evaluated againstskills.sh.jsonbefore writing anything. - Someone wants to compile a batch of wiki pages into
SKILL.mdfiles and needs the propose-only vs. write-mode boundary made explicit first. - Do not use this skill to answer a one-off question or to publish domain facts as-is; use it only when the target output is a procedural, triggerable skill.
Inputs
- Candidate wiki pages or folders.
- Existing
skills/directory. skills.sh.jsonif present.- Target agent(s).
- Desired mode: propose-only or write skill files.
Procedure
1. Find procedural candidates
Search for pages containing:
- step-by-step workflows;
- repeated troubleshooting routines;
- runbooks;
- checklists;
- project conventions;
- saved query answers that are reused;
- scripts or command sequences.
2. Separate procedure from knowledge
Keep in the skill:
- trigger conditions;
- required inputs;
- deterministic procedure;
- output contract;
- safety gates.
Keep in the wiki or references:
- domain facts;
- long examples;
- historical background;
- volatile current-state claims;
- source excerpts.
3. Design skill boundary
For each candidate, propose:
| Candidate | Skill name | Trigger | Inputs | Risks | Evidence pages |
|---|---|---|---|---|---|
Use one skill per coherent user intent. Do not create one giant skill.
4. Generate SKILL.md
Create valid frontmatter:
---
name: example-skill
description: Trigger-rich description.
license: MIT
compatibility: Designed for Agent Skills-compatible coding agents.
metadata:
author: po4yka
version: "0.1.0"
install_scope: self-contained
---
Set install_scope: self-contained when the skill bundles everything it references; use pack-install-only and explain the requirement in compatibility when it depends on a full repository install.
Add links back to source wiki pages in a ## Source knowledge section when useful.
5. Update package metadata and validate
If writing files, update skills.sh.json, then validate the generated skills with the bundled read-only auditor:
node scripts/audit-skills.mjs <path-to-generated-skills>
Fix every reported error before delivering; treat warnings as review items. Do not declare a generated skill valid without running the script.
Output
Deliver the summary in chat; the persisted artifacts are the generated skill directories and the updated skills.sh.json.
## Candidate skills
## Generated skills
## Source knowledge used
## Validation status
## Review required
Safety gates
- Do not compile private or sensitive wiki content into distributable skills without approval.
- Do not embed large volatile knowledge dumps in
SKILL.md. - Do not create a skill from a weak or unreviewed page without marking it draft.
- Do not overwrite existing skills without diff review.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |