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 skills add nxtg-ai/forge-plugin --skill agent-developmentgit clone --depth 1 https://github.com/nxtg-ai/forge-pluginWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/nxtg-ai/forge-plugin/agent-development)<a href="https://agentmods.dev/skills/nxtg-ai/forge-plugin/agent-development"><img src="https://agentmods.dev/badge/skills/nxtg-ai/forge-plugin/agent-development/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/nxtg-ai/forge-plugin/agent-development"><img src="https://agentmods.dev/badge/skills/nxtg-ai/forge-plugin/agent-development.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What 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 5.1 | $0.00083 | $0.02616 |
| Opus 5.5 | $0.00033 | $0.01046 |
| Sonnet 5.5 | $0.00017 | $0.00523 |
| Haiku 4.5 | $0.00008 | $0.00262 |
Grade A, and why
Agent Development 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 1mo ago.
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 — 220 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent Development Guide
How to author a Claude Code subagent that loads, auto-delegates, and runs with the right tools.
Ground every new agent against the 33 real agents in this plugin (agents/*.md) — copy a close
match and adapt, don't start from a blank file.
Where agents live
- Plugin agents (shipped):
plugins/nxtg-forge/agents/<name>.md. - Project-local agents (a user's repo):
.claude/agents/<name>.md. - Personal agents (all your projects):
~/.claude/agents/<name>.md.
An agent is a single markdown file: YAML frontmatter (config) + body (the system prompt).
Frontmatter — valid fields only
---
name: builder # REQUIRED. lowercase + digits + hyphens, ≤64 chars. No uppercase, no underscores.
description: | # REQUIRED. When Claude should delegate here. Add <example> blocks.
Use this agent when ... This includes: ...
<example>
Context: ...
user: "..."
assistant: "I'll use the Task tool to launch the builder agent to ..."
<commentary>Why builder is the right pick.</commentary>
</example>
model: sonnet # optional. sonnet | opus | haiku. Omit to inherit the main thread's model.
color: green # optional. EXACTLY one of: purple cyan green orange blue red
tools: Glob, Grep, Read, Write, Edit, Bash, TodoWrite # optional allowlist. Omit Task for leaf workers.
isolation: worktree # optional. Own git worktree for conflict-free parallel file edits.
memory: project # optional. project | user | local. Persists MEMORY.md across sessions.
skills: nxtg-forge:parallel-execution # optional. Preloads FULL skill text at startup (token cost).
disallowedTools: WebFetch # optional denylist, complements `tools`.
permissionMode: acceptEdits # optional. default | acceptEdits | plan | dontAsk | bypassPermissions
---
Field notes
- name must match the delegate name Claude uses ("launch the builder agent"). It is also the discovery key — invalid casing/underscores make the agent silently not load.
- description is the routing engine, not documentation — see below.
- model:
sonnetfor most specialists;opusfor orchestration/high-stakes judgment (orchestrator,ceo-loop);haikuonly for cheap mechanical work. Omit to inherit. - tools: allowlist. A leaf worker that must never spawn subagents omits
Task. - isolation: worktree: use for agents that write files in parallel with others (
builder,testing) so edits land on separate branches instead of colliding. - memory: project: use for agents that must remember across sessions (
learning). Writes a git-trackedMEMORY.md. - skills: preloads the entire skill body into the agent's context at startup — real routing cost, so only preload skills the agent uses every run.
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.
- 1mo ago First seen · 220 lines · 83 tokens per session scan A 4efe2df0b065
Agent Development is a skill published in the GitHub repository nxtg-ai/forge-plugin (6 stars, last pushed 4d ago), licensed MIT. It adds 83 tokens to every session and 2,616 once invoked, about $0.0003 per session on Opus 5.5. 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 2026-09-03.
Other skills, from other repositories
rag-retrieval
Retrieval-Augmented Generation patterns for grounded LLM responses. Use when building RAG pipelines, embedding documents, implementing hybrid search, contextual retrieval, HyDE, agentic RAG, multimodal RAG, query decomposition, reranking, or pgvector search.
error-analysis
Evals-first error analysis for LLM apps: clusters real Langfuse or JSONL traces into a human-confirmed failure taxonomy with counts, then recommends binary pass/fail evals for recurring named modes. Use to learn what to measure before writing evals. Not for CI failures.
dev
One-command dev loop boot. Spins up portless (named HTTPS subdomain), emulate (stateful API mocks), the project's dev server, and an agent-browser session, all keyed to the current git branch. Use when starting a feature branch, switching worktrees, or returning to a project after a break. Skips silently with install…
testing-llm
LLM and AI testing patterns — mock responses, evaluation with DeepEval/RAGAS, structured output validation, and agentic test patterns (generator, healer, planner). Use when testing AI features, validating LLM outputs, or building evaluation pipelines.
accessibility
Accessibility patterns for WCAG 2.2 compliance, keyboard focus management, React Aria component patterns, cognitive inclusion, native HTML-first philosophy, and user preference honoring. Use when implementing screen reader support, keyboard navigation, ARIA patterns, focus traps, accessible component libraries…
manage-schemas
Creates, views, edits, deletes, and validates note schemas for the MCP Task Orchestrator in .taskorchestrator/config.yaml — the templates that define which notes agents must fill at each workflow phase. Also recommends schema designs from a library of workflow patterns (autonomous coding loops, spec-driven teams…