Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/sharpdeveye/maestro/specializenpx skills add sharpdeveye/maestro --skill specializegit clone --depth 1 https://github.com/sharpdeveye/maestroWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00031 | $0.00499 |
| Opus 5 | $0.00015 | $0.00249 |
| Sonnet 5 | $0.00006 | $0.00100 |
| Haiku 4.5 | $0.00003 | $0.00050 |
Grade A, and why
specialize scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 63 lines — stays where its author put it; the contents beside it link to each section on GitHub.
MANDATORY PREPARATION
Invoke /agent-workflow — it contains workflow principles, anti-patterns, and the Context Gathering Protocol. Follow the protocol before proceeding — if no workflow context exists yet, you MUST run /teach-maestro first.
Transform a general-purpose workflow into a domain expert.
Step 1: Domain Discovery
- Terminology: What domain-specific language must the agent use?
- Regulations: What compliance requirements apply? (HIPAA, SOC2, GDPR)
- Standards: What industry standards govern output format or quality?
- Expert expectations: What would a domain expert check first?
- Common errors: What mistakes would a non-expert make?
Step 2: Specialize Prompts
## Generic: You are an assistant that analyzes documents.
## Specialized (legal): You are a senior legal analyst specializing in contract review.
You understand common law jurisdictions, standard contract clauses, and the
difference between representations and warranties. Always caveat that this
is not legal advice.
Step 3: Specialized Evaluation
| Domain | Evaluation Criteria |
|---|---|
| Legal | Clause completeness, regulatory compliance, jurisdiction accuracy |
| Medical | Clinical accuracy, guideline adherence, contraindication checks |
| Financial | Calculation accuracy, regulatory disclosure, risk assessment |
| Code | Test coverage, security vulnerabilities, performance |
| Customer Support | Tone, escalation accuracy, resolution completeness |
Step 4: Domain Guardrails
- Legal: "Not legal advice" disclaimer, jurisdiction limitations
- Medical: "Not medical advice" disclaimer, emergency detection
- Financial: Regulatory disclosures, suitability warnings
- Code: Security scanning, dependency vulnerability checks
Recommended Next Step
After specialization, run /evaluate with domain-specific scenarios, then /guard to add domain-appropriate safety guardrails.
NEVER:
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 63 lines · 31 tokens per session scan A 65d7a876a7b0
specialize is a skill published in the GitHub repository sharpdeveye/maestro (415 stars, last pushed 4mo ago), licensed MIT. It adds 31 tokens to every session and 499 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
agent-self-scheduling
Schedule AI agent runs with cron, loops, or external clocks while avoiding unsafe tight autonomous timers.
skill-creator
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
agnt-plugin-builder
End-to-end workflow for creating, building, installing, and hot-reloading AGNT plugins entirely from chat. Use this skill whenever the user asks to 'build a plugin', 'create an AGNT plugin', 'add a new tool to AGNT', 'integrate X with AGNT' (where X is an API or service), 'make a plugin for [service]', or wants to…
annie-universal-api-orchestrator
Use AGNT's stored OAuth tokens and API keys to call ANY third-party API directly from the orchestrator, without building a tool or plugin first. Use this skill whenever the user asks you to "do something with my GitHub / Gmail / Drive / Slack / Notion / Stripe / Shopify / Discord / Linear / Jira / Vercel / Netlify /…
frontend-slides
Create stunning, animation-rich HTML presentations from scratch or by converting PowerPoint files. Use when the user wants to build a presentation, convert a PPT/PPTX to web, or create slides for a talk/pitch. Helps non-designers discover their aesthetic through visual exploration rather than abstract choices.
image-to-cinematic-video
Turn a prompt or an existing image into a polished multi-scene cinematic short video using Seedance for clip generation, uguu.se for file hosting, and FFmpeg for last-frame extraction and crossfade stitching. Supports two modes — PARALLEL (multiple scenes from the same reference image, concurrent, 3 min total) and…