Special Skills — Productivity skill for Claude Code
Graph engineering and bounded-behavior plugins for coding agents: validated workflow graphs, loop limits, receipts, and prompt-injection defense for Claude Code, Codex, and any Agent Skills host.
How to install Special Skills
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open MTEnt/special-skills and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Special Skills does
Graph engineering and bounded-behavior plugins for coding agents: validated workflow graphs, loop limits, receipts, and prompt-injection defense for Claude Code, Codex, and any Agent Skills host.
Alternatives in Productivity
- Nosqli — NoSQL injection scanner (MongoDB/Mongoose/operator-injection DBs) — auth bypass via $ne/$gt operators, bracket 4.5k ★
- Run Discovery — Workflow recipe — go from a vague opportunity to validated insight and a prioritised next step by chaining 4 s 1.3k ★
- Build In Public — Build-in-Public — generate posts from your engineering work, anonymize, and schedule via Buffer 705 ★
README
Special Skills
**Graph engineering and bounded-behavior plugins for coding agents.**
Special Skills is a maintained collection of agent plugins for Claude Code, Codex, and any agent that reads the [Agent Skills](https://agentskills.io) format. Every package makes agent behavior more reliable, bounded, inspectable, or repeatable, and every claim of that kind is backed by something you can run: a hook, a validator, a test, or an eval.
Start here
| If you want to | Install |
|---|---|
| Decide whether a workflow needs a graph at all, then design, validate, compile, audit, or evolve one as a versioned artifact | `graph-engineering` |
| Stop an agent from looping on the same failed fix and force an evidence-backed handoff | `anti-loop` |
| Get a receipt after every change and a verification gate before "done" | `cleancoding` |
| Treat fetched pages, files, and tool output as data instead of instructions | `untrusted-content` |
`graph-engineering` covers execution graphs: agents, tools, typed state, edges, checkpoints, and approvals as one validated artifact. It is not a knowledge-graph or GraphRAG tool; it says so and routes those requests elsewhere.
Install
Any agent that supports Agent Skills (Claude Code, Codex, Cursor, Copilot, Gemini CLI, and others):
npx skills add MTEnt/special-skills --skill graph-engineeringPlugins that ship lifecycle hooks (`cleancoding`, `anti-loop`, `anti-amnesia`, `temporal-tasks`) need the plugin route so the hooks install with the skill. Add the marketplace once, then install what you need.
Claude Code:
/plugin marketplace add MTEnt/special-skills
/plugin install graph-engineering@special-skillsCodex:
codex plugin marketplace add MTEnt/special-skills
codex plugin add graph-engineering@special-skillsEvery plugin installs the same way; substitute the name from the t
Related Skills
Lasso Claude-hooks
Prompt injection defense hooks: scans files, web fetches, and command output in real-time. Detects 50+ injecti
Escalate
Reframe a blocked decision or explicitly delegate a bounded task using current host capabilities.
Feature Builder
Engineering Workflow Directive — Feature Implementation & Integrations. Implements a feature-request epic or s
Atomic
Run a task through recon → plan → parallel build → hostile critic → bounded loop.
Slm Loop
Run a task as a bounded, gate-verified loop backed by SuperLocalMemory — iterate until an independent gate pas
Spectre
Spectre is a battle-tested 6 stage agentic engineering workflow for building complex, production-ready feature
Related Agents
Generative UI Agent
Generate HTML dashboard from financial data and graphs. Use after graph generation.
Dependency Graph Architect
Use this agent when designing or reviewing dependency graph structures, adapter composition patterns, port com
Security Architecture
Boucle implements a defense-in-depth security model to protect against prompt injection attacks and maintain t
