KS
Killer-Skills

oracle — how to use oracle how to use oracle, what is oracle, oracle alternative, oracle vs matrixshader, oracle install, oracle setup guide, matrix rain windows terminal, ai agent skills for windows terminal, automating faq questions

v1.0.0
GitHub

About this Skill

Ideal for AI Agents like Claude and AutoGPT requiring advanced FAQ analysis and codebase reading capabilities. Oracle is a Matrix rain AI agent skill for Windows Terminal that automates FAQ question processing and answering.

Features

Fetches pending questions via GET /api/faq?status=pending with Bearer auth
Reads codebase files to build answer context
Classifies questions into CAN ANSWER, NEEDS HUMAN, or SPAM categories
Requires user approval before executing each action
Supports drafting accurate answers using the codebase
Identifies spam for cleanup

# Core Topics

matrixshader matrixshader
[0]
[0]
Updated: 3/1/2026

Quality Score

Top 5%
42
Excellent
Based on code quality & docs
Installation
SYS Universal Install (Auto-Detect)
Cursor IDE Windsurf IDE VS Code IDE
> npx killer-skills add matrixshader/matrix-shader/oracle

Agent Capability Analysis

The oracle MCP Server by matrixshader is an open-source Categories.community integration for Claude and other AI agents, enabling seamless task automation and capability expansion. Optimized for how to use oracle, what is oracle, oracle alternative.

Ideal Agent Persona

Ideal for AI Agents like Claude and AutoGPT requiring advanced FAQ analysis and codebase reading capabilities.

Core Value

Empowers agents to draft accurate answers by reading codebase files, classify questions into actionable categories, and identify spam using protocols like Bearer auth and APIs such as GET /api/faq, thereby streamlining support page management.

Capabilities Granted for oracle MCP Server

Automating FAQ question processing
Generating accurate answers from codebase context
Flagging business questions for human review
Identifying spam for cleanup

! Prerequisites & Limits

  • Requires user approval for each action
  • Needs access to codebase files
  • Dependent on MatrixShader support page API
Project
SKILL.md
5.2 KB
.cursorrules
1.2 KB
package.json
240 B
Ready
UTF-8

# Tags

[No tags]
SKILL.md
Readonly
<objective> Process all pending FAQ questions submitted via the MatrixShader support page. Reads the codebase to draft accurate answers for questions Claude can handle, flags business/roadmap questions for the human, and identifies spam for cleanup. Every action requires user approval before executing. </objective>

<quick_start>

  1. Fetch pending questions: GET /api/faq?status=pending with Bearer auth
  2. Read codebase files to build answer context
  3. Classify each question into CAN ANSWER, NEEDS HUMAN, or JUNK
  4. Present all results grouped by bucket with draft answers
  5. User approves/edits/skips each — then execute API calls </quick_start>

<essential_principles>

  • Never fabricate answers. Only answer from codebase knowledge, README, and docs. If unsure, classify as NEEDS HUMAN.
  • Never publish without approval. Always show the draft and wait for the user to confirm.
  • Keep answers concise. 1-3 sentences. Helpful and direct, matching the Matrix operator voice (but not over-the-top).
  • Batch efficiently. Show all CAN ANSWER questions together so the user can approve in bulk or review individually. </essential_principles>
<context> **API base:** `https://matrixshader.com/api/faq`

Auth: Authorization: Bearer $DASHBOARD_PASSWORD header (env var). If not set, ask the user for it.

Endpoints:

  • GET /api/faq?status=pending — fetch pending questions (auth required)
  • PATCH /api/faq — update a question (auth required). Body: { "id", "action", ... }
    • action: "publish" — requires answer field, optional category
    • action: "dismiss" — marks as dismissed
    • action: "reopen" — returns to pending
    • action: "update" — partial update (answer, category, tags)
  • DELETE /api/faq?id=xxx — permanently delete (auth required)

Valid categories: general, installation, usage, licensing, compatibility, troubleshooting </context>

<process> **Step 1: Authenticate and fetch**

Run scripts/fetch-pending.sh to get all pending questions. If DASHBOARD_PASSWORD is not in environment, ask the user for it.

If zero pending questions, say "The Oracle sees no unanswered questions." and stop.

Step 2: Display summary

Show a numbered overview:

The Oracle: {N} pending questions

#1 [{category}] "{question}" — {email} ({date})
#2 [{category}] "{question}" — {email} ({date})
...

Step 3: Build answer context

Read these codebase files to prepare for drafting answers:

  • CLAUDE.md — architecture overview, key mechanisms
  • README.md — user-facing feature list, install instructions
  • Website/index.html — product page feature descriptions
  • Website/redpill/index.html — Red Pill purchase page details
  • matrix_control.ps1 — control panel features and hotkeys
  • matrix_setup.ps1 — setup wizard flow
  • Website/privacy/index.html and Website/terms/index.html — policy questions

Search additional files as needed for specific questions (shader files, C# source, etc.).

Step 4: Classify each question

For each pending question:

  1. Check if the category is correct — note any corrections needed
  2. Determine which bucket it belongs in:
BucketCriteriaExamples
CAN ANSWERAnswerable from codebase/docs"How do I change colors?", "Does it work on Windows 10?"
NEEDS HUMANRequires business judgment, account info, or roadmap commitment"Can I get a discount?", "When will you support macOS?", "I want a refund"
JUNKGibberish, test data, spam, not a real question"asdfasdf", "test123"

Step 5: Present results

Group by bucket and show:

For CAN ANSWER — show question, category correction (if any), and draft answer For NEEDS HUMAN — show question and explain why only the human can answer it For JUNK — show question and recommend dismiss or delete

Ask the user how to proceed. Accept bulk commands ("publish all", "looks good") or individual review.

Step 6: Execute approved actions

For each approved action, run the appropriate API call using scripts/faq-action.sh:

  • Publish: scripts/faq-action.sh publish <id> "<answer>" [category]
  • Dismiss: scripts/faq-action.sh dismiss <id>
  • Delete: scripts/faq-action.sh delete <id>
  • Update category: scripts/faq-action.sh update <id> [category]

Step 7: Summary

Show final tally:

Oracle Complete:
  Published: {N}
  Dismissed: {N}
  Deleted:   {N}
  Skipped:   {N} (still pending)
</process>

<anti_patterns>

  • Guessing answers — If you're not confident from codebase sources, mark NEEDS HUMAN. Wrong answers on a public FAQ are worse than no answer.
  • Publishing without approval — Every publish must be explicitly confirmed by the user.
  • Over-theming answers — Answers should be helpful and clear. Don't force Matrix references into every response.
  • Skipping category correction — If a question is miscategorized, fix it as part of the publish/update call. </anti_patterns>

<success_criteria> Triage is complete when:

  • All pending questions have been classified into buckets
  • User has reviewed and decided on each question (publish, dismiss, delete, or skip)
  • All approved API calls have been executed successfully
  • Final summary shows the tally of actions taken </success_criteria>

Related Skills

Looking for an alternative to oracle or building a Categories.community AI Agent? Explore these related open-source MCP Servers.

View All

widget-generator

Logo of f
f

widget-generator is an open-source AI agent skill for creating widget plugins that are injected into prompt feeds on prompts.chat. It supports two rendering modes: standard prompt widgets using default PromptCard styling and custom render widgets built as full React components.

149.6k
0
Design

chat-sdk

Logo of lobehub
lobehub

chat-sdk is a unified TypeScript SDK for building chat bots across multiple platforms, providing a single interface for deploying bot logic.

73.0k
0
Communication

zustand

Logo of lobehub
lobehub

The ultimate space for work and life — to find, build, and collaborate with agent teammates that grow with you. We are taking agent harness to the next level — enabling multi-agent collaboration, effortless agent team design, and introducing agents as the unit of work interaction.

72.8k
0
Communication

data-fetching

Logo of lobehub
lobehub

The ultimate space for work and life — to find, build, and collaborate with agent teammates that grow with you. We are taking agent harness to the next level — enabling multi-agent collaboration, effortless agent team design, and introducing agents as the unit of work interaction.

72.8k
0
Communication