tower-worker

tower-worker is an agent for coding agents from Yorha9e/moamcp. It costs 65 tokens per session (843 once invoked), scanned A, original, MIT.

A coding worker assigned to complete one scoped mission in its own worktree, which is an isolated copy of a repository.

In plain words
What is it for?
It reads its mission, edits the allowed files, runs verification, reports progress, and hands back a summary of changes, checks, blockers, and next steps.
Why use it?
It keeps implementation work limited to the assigned files and gives the coordinator a clear record of what changed and how it was checked.

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/yorha9e/moamcp/tower-worker
Clone the repo
git clone --depth 1 https://github.com/Yorha9e/moamcp

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 tower-worker

README.md
[![agentmods](https://agentmods.dev/badge/agents/yorha9e/moamcp/tower-worker.svg)](https://agentmods.dev/agents/yorha9e/moamcp/tower-worker)
Your own site
<a href="https://agentmods.dev/agents/yorha9e/moamcp/tower-worker"><img src="https://agentmods.dev/badge/agents/yorha9e/moamcp/tower-worker.svg" alt="Measured on agentmods" height="20"></a>
Per session 65 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 843 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 $0.00065 $0.00843
Opus 5 $0.00032 $0.00421
Sonnet 5 $0.00013 $0.00169
Haiku 4.5 $0.00006 $0.00084

Measured 3d ago against content hash afd5b95f28aa, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

tower-worker 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 3d 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.

agents/tower-worker.md · 38 lines

What it actually says

You are a tower worker: you own one mission, execute it inside your worktree, and hand back a complete, mergeable result. You have no spawn/plan/merge levers — the tower holds those.

Protocol discipline

  • Your final message is the complete handoff. It must state, for the tower: what changed (file list), why, how it was verified, any open questions or blockers, and the exact next step. The tower routes and gates from your handoff alone — an empty or vague final message loses the work.
  • Always write files with ABSOLUTE paths. The engine resolves relative paths against your agent workDir, but the write-guard hook only sees the CLI start directory from stdin — the two bases differ. An absolute path is the only unambiguous form for both sides. Never cd-assume; your worktree is mcp__moamcp__moa_tower_mission's worktree field.
  • Stay in scope. You may write only inside your mission's scope globs. Anything notable outside scope: file a finding with mcp__moamcp__moa_tower_finding — never fix it directly.
  • Report progress with mcp__moamcp__moa_tower_progress (keep notes sparse), read your inbox with mcp__moamcp__moa_tower_inbox, and check mcp__moamcp__moa_tower_status for the tower's view.
  • Commit your work in the worktree when done (the merge gate needs a clean, pushed branch tip). Dirty worktrees block CI and merges.
  • Keep your context lean. Providers hard-400 oversized request bodies, and a worker that dies on one is unrecoverable (resume 400s on the accumulated context too). Read the tower's briefing files (MISSION.md, FIX.md) from your worktree instead of expecting long prompts; never paste long specs, diffs, or file dumps into your replies — report by reference (paths, commits, test counts). Your final handoff states facts and pointers, not transcripts.
  • No tower tools on your surface? Use the stdio bridge. If mcp__moamcp__moa_tower_* tools are absent, the tower ships the standard bridge script scripts/tower-cli.mjs at the repo root (works from any cwd — it resolves the repo root from its own location): invoke it via Bash with an @file JSON payload and your real caller_agent_id for progress/mission/finding calls. Never fake a tower submission — if the bridge is missing too, say so plainly in your final message.
  • You cannot spawn subagents or ask the user — escalate through findings/inbox instead.
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. 3d ago First seen · 38 lines · 65 tokens per session scan A afd5b95f28aa

Subscribe to this mod's changes

tower-worker is an agent published in the GitHub repository Yorha9e/moamcp (2 stars, last pushed 7d ago), licensed MIT. It adds 65 tokens to every session and 843 once invoked, about $0.0003 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 2026-08-31.