Sandcastle is a TypeScript library that runs AI coding agents inside isolated sandboxes and merges their branch changes back into a project. It is used to run agents in parallel, build review pipelines, or coordinate agent work through providers such as Docker, Podman, or Vercel.
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 agents/mattpocock/sandcastle/adding-an-agent-providergit clone --depth 1 https://github.com/mattpocock/sandcastleWrote 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/agents/mattpocock/sandcastle/adding-an-agent-provider)<a href="https://agentmods.dev/agents/mattpocock/sandcastle/adding-an-agent-provider"><img src="https://agentmods.dev/badge/agents/mattpocock/sandcastle/adding-an-agent-provider.svg" alt="Measured on agentmods" 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 | $0.00000 | $0.03310 |
| Opus 5 | $0.00000 | $0.01655 |
| Sonnet 5 | $0.00000 | $0.00662 |
| Haiku 4.5 | $0.00000 | $0.00331 |
Grade A, and why
adding-an-agent-provider 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 5d 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 — 175 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Adding an agent provider
This document is for contributors adding support for a new agent (e.g. Claude Code, Codex, gemini-cli) as a built-in agent provider. It covers:
- Evaluating a new agent — the questionnaire used to decide whether an agent's CLI can be supported.
- The
AgentProviderinterface — what you implement. - Scaffold integration — what
sandcastle initneeds to offer the agent. - Implementation checklist — every file to touch.
For terminology (agent, agent provider, sandbox, etc.), see CONTEXT.md.
Evaluating a new agent
Before implementing, confirm the agent's CLI satisfies the must-haves below. If a must-have is missing, the agent likely cannot be supported until its CLI changes.
Must-have CLI capabilities
- Non-interactive run mode. A flag (often
--print,exec,run) that takes a prompt, runs the agent to completion, and exits. Without this, Sandcastle cannot drive the agent unattended. - Prompt input via argv or stdin. stdin is strongly preferred — Linux caps
argvat ~128 KB, which large prompts blow past. - Auto-approval / bypass-permissions flag. The agent runs inside a sandbox, so any "are you sure?" prompts will hang it. Examples: Claude Code's
--dangerously-skip-permissions, Codex's--dangerously-bypass-approvals-and-sandbox. - Model selection. A flag that picks the model (e.g.
--model,-m). - Env-based authentication. Auth via environment variables (API key) — no interactive login required at first run.
- Exit codes reflect success/failure. Non-zero on error so the orchestrator can detect failures.
Must-have output capabilities
- Streaming output. The CLI must stream output to stdout as the agent runs; we surface it live in
Display. - Structured (JSON) stream events. Required — line-delimited JSON (one event per line). This is what
parseStreamLineinAgentProvider.tsconsumes; without it, tool calls and partial text cannot render in the UI.
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.
- 5d ago First seen · 175 lines · 0 tokens per session scan A e0a8673e7108
adding-an-agent-provider is an agent published in the GitHub repository mattpocock/sandcastle (7,830 stars, last pushed 2mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 3,310 tokens. 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 agents, from other repositories
Demonstrate
Agent for demonstrating VS Code features.
playwright-test-generator
Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.
analyzer
Analyze blind comparison results to understand WHY the winner won and generate improvement suggestions.
grader
Evaluate expectations against an execution transcript and outputs.
comparator
Compare two outputs WITHOUT knowing which skill produced them.
.NET-Notebook-Migration-Agent
Expert .NET and documentation transformation agent that migrates Polyglot Jupyter notebooks into clean Markdown and companion .NET sample code.