E-phase-5-sandbox

E-phase-5-sandbox is an agent for coding agents from darkiceinteractive/mcp-conductor. It costs 0 tokens per session (494 once invoked), scanned A, original, MIT.

A development task for adding search, summarising, budget, and skill-loading features to an MCP-based coding-agent system. MCP is a way for an agent to use external tools through a standard interface.

In plain words
What is it for?
Implementing compact, summarise, delta, budget, and findTool helpers; loading skills from a configured directory; watching for skill changes; and caching local search embeddings.
Why use it?
It is intended to reduce unnecessary back-and-forth with tools and help the agent find relevant tools and skills locally.

Agent

Install

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.

agentmods
npx agentmods add agents/darkiceinteractive/mcp-conductor/e-phase-5-sandbox
Clone the repo
git clone --depth 1 https://github.com/darkiceinteractive/mcp-conductor

Wrote 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.

agentmods badge for E-phase-5-sandbox

README.md
[![agentmods](https://agentmods.dev/badge/agents/darkiceinteractive/mcp-conductor/e-phase-5-sandbox.svg)](https://agentmods.dev/agents/darkiceinteractive/mcp-conductor/e-phase-5-sandbox)
Your own site
<a href="https://agentmods.dev/agents/darkiceinteractive/mcp-conductor/e-phase-5-sandbox"><img src="https://agentmods.dev/badge/agents/darkiceinteractive/mcp-conductor/e-phase-5-sandbox.svg" alt="Measured on agentmods" height="20"></a>
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 494 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00000 $0.00494
Opus 5 $0.00000 $0.00247
Sonnet 5 $0.00000 $0.00099
Haiku 4.5 $0.00000 $0.00049

Measured 4d ago against content hash 22562647f6fc, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

E-phase-5-sandbox 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 4d 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.

_plans/v3-enhancements/agents/E-phase-5-sandbox.md · 49 lines

What it actually says

Agent E — PRD Phase 5 Sandbox capabilities

You add mcp.compact/summarize/delta/budget/findTool + skills wiring + vector tool index.

Setup

pwd  # /mcp-executor-darkice-worktrees/E-phase-5
git branch --show-current  # feature/v3-phase-5
git fetch origin
# Need both Phase 1 AND Phase 4
git rebase origin/feature/v3-phase-0-1
git rebase origin/feature/v3-phase-4
npm install && npm run test:run

Read PRD §5 Phase 5 (lines ~835–986).

Append start checkpoint to STATUS.md.

Scope

Build per PRD: src/runtime/helpers/, src/runtime/findtool/, extend src/skills/. Plug helpers into Agent D's worker preload hook.

findTool uses local MiniLM-L6 ONNX (from @xenova/transformers or onnxruntime-node). Lazy initialise on first call. Cache embeddings to disk under ~/.mcp-conductor/embeddings/ keyed by registry revision.

Skills wiring: read process.env.CLAUDE_SKILLS_DARKICE first; fall back to ./skills. Hot reload on directory change (existing skills engine has the watcher).

Acceptance

PRD §5 Phase 5 "Acceptance criteria" — all helpers must be zero-roundtrip, findTool returns relevant tool in top 3.

Commit pattern

feat(v3-phase-5): mcp.compact field selection + array truncation
feat(v3-phase-5): mcp.summarize heuristic with style modes
feat(v3-phase-5): mcp.delta wrapping cache.delta
feat(v3-phase-5): mcp.budget auto-trim with BudgetExceededError
feat(v3-phase-5): findTool vector index (MiniLM-L6 local ONNX)
feat(v3-phase-5): skills engine wired to $CLAUDE_SKILLS_DARKICE
feat(v3-phase-5): preload helpers into worker pool
test(v3-phase-5): full helpers/findtool/skills test suite

PR + stop

gh pr create --base feature/v3 --title "v3 Phase 5: Sandbox capabilities". STATUS.md.

Changes

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.

  1. 4d ago First seen · 49 lines · 0 tokens per session scan A 22562647f6fc

Subscribe to this mod's changes

E-phase-5-sandbox is an agent published in the GitHub repository darkiceinteractive/mcp-conductor (0 stars, last pushed 2mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 494 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 2026-08-31.