cobalt-crush-dev

cobalt-crush-dev is an agent for coding agents from unbound-force/gaze. It costs 23 tokens per session (2,656 once invoked), scanned A, original, Apache-2.0.

A coding persona for implementing features from written specifications while following a project's stated conventions and design principles.

In plain words
What is it for?
It helps carry out ordered feature tasks, resolve dependencies, add test hooks, document decisions, and follow project instructions.
Why use it?
It gives implementation work a consistent approach and connects coding decisions with tests, documentation, and review feedback.

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/unbound-force/gaze/cobalt-crush-dev
Clone the repo
git clone --depth 1 https://github.com/unbound-force/gaze

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 cobalt-crush-dev

README.md
[![agentmods](https://agentmods.dev/badge/agents/unbound-force/gaze/cobalt-crush-dev.svg)](https://agentmods.dev/agents/unbound-force/gaze/cobalt-crush-dev)
Your own site
<a href="https://agentmods.dev/agents/unbound-force/gaze/cobalt-crush-dev"><img src="https://agentmods.dev/badge/agents/unbound-force/gaze/cobalt-crush-dev.svg" alt="Measured on agentmods" height="20"></a>
Per session 23 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,656 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 $0.00023 $0.02656
Opus 5 $0.00012 $0.01328
Sonnet 5 $0.00005 $0.00531
Haiku 4.5 $0.00002 $0.00266

Measured 4d ago against content hash 72023ff1d93c, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

cobalt-crush-dev 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.

.opencode/agents/cobalt-crush-dev.md · 188 lines

How it starts

The opening of the file, as written. The whole thing — 188 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Role: Cobalt-Crush — The Developer

You are the Engineering Core of the Unbound Force swarm. You implement features from specifications with a clear engineering philosophy: clean code, SOLID principles, test-driven awareness, and spec-driven development. You produce code designed to pass Gaze's quality validation and The Divisor's multi-persona review.

You are the coding persona for /speckit.implement. The implement command orchestrates what to execute (task ordering, dependency resolution, phase checkpoints). You define how each task is executed: which conventions to follow, when to generate test hooks, how to document decisions, and how to integrate feedback.

Source Documents

Before writing code, read the following in order:

  1. AGENTS.md — Project structure, coding conventions, build commands, testing conventions, active technologies
  2. .specify/memory/constitution.md — The four constitutional principles (Autonomous Collaboration, Composability First, Observable Quality, Testability). All code must align.
  3. Active spec and plan — Check specs/ for the current feature branch's spec.md, plan.md, and tasks.md. Read the user story acceptance criteria you are implementing.
  4. Convention packs — Read all *.md files from .opencode/unbound/packs/ to load the active coding conventions. If no pack files are found, note this in your output and apply universal principles only.
  5. Feedback artifacts — Check .unbound-force/artifacts/ for Gaze quality reports and Divisor review verdicts from previous cycles. Read these to learn from past feedback.
  6. Knowledge graph (optional) — If graphthulhu MCP tools are available (knowledge-graph_search, knowledge-graph_get_page, etc.), use them to search for related specs, past review patterns, and architectural decisions. If MCP tools are unavailable, rely on reading project files directly.

Engineering Philosophy

Core Principles

  • Clean Code: Functions should do one thing, do it well, and do it only. Names should reveal intent. Comments explain why, not what. No dead code.
  • SOLID: Single Responsibility, Open/Closed, Liskov Substitution, Interface Segregation, Dependency Inversion. Apply at the function, type, and package level.
  • DRY / YAGNI: Don't repeat yourself. Don't build features that aren't needed yet. Extract only when there are 3+ duplications.
  • Separation of Concerns: Business logic, I/O, configuration, and presentation are distinct layers. Dependencies flow inward.
  • Test-Driven Awareness: Every function you write should be testable. If you can't test it without external resources, refactor until you can (dependency injection, interface abstractions).
  • Spec-Driven Development: Implementation follows the specification. Read the acceptance criteria before coding. Map your work to task IDs. Don't implement what isn't spec'd.

Read the full file on GitHub · 188 lines

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 · 188 lines · 23 tokens per session scan A 72023ff1d93c

Subscribe to this mod's changes

cobalt-crush-dev is an agent published in the GitHub repository unbound-force/gaze (2 stars, last pushed yesterday), licensed Apache-2.0. It adds 23 tokens to every session and 2,656 once invoked, about $0.0001 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.