deep-worker

deep-worker is an agent for Claude Code from kvdm-co-pilot/create-cmp. It costs 78 tokens per session (607 once invoked), scanned A, original, MIT.

A single-task coding worker for careful, evidence-based engineering work. It handles jobs such as falsification checks, architecture decision records, broad refactors, and adversarial reviews.

In plain words
What is it for?
Use it for one isolated task that needs thorough investigation, honest reporting, or changes across many files.
Why use it?
It reduces the risk of unsupported claims, untested conclusions, and changes made only to satisfy a failing check. It requires findings to be based on files or observed results and encourages attempts to disprove them.

Agent for Claude Code

Written for Claude Code: effort in frontmatter. Also seen: model in frontmatter.

Part of the create-cmp plugin — 15 skills, 2 agents, 4 hooks, 1 MCP server shipped together

Good fit Use it for one isolated task that needs thorough investigation, honest reporting, or changes across many files.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/kvdm-co-pilot/create-cmp/deep-worker
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.

Clone the repo
git clone --depth 1 https://github.com/kvdm-co-pilot/create-cmp

Made for: Claude Code.

Or install create-cmp, the plugin that ships this one along with the rest of its 15 skills, 2 agents, 4 hooks, 1 MCP server.

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 deep-worker

README.md
[![agentmods](https://agentmods.dev/badge/agents/kvdm-co-pilot/create-cmp/deep-worker.svg)](https://agentmods.dev/agents/kvdm-co-pilot/create-cmp/deep-worker)
Your own site
<a href="https://agentmods.dev/agents/kvdm-co-pilot/create-cmp/deep-worker"><img src="https://agentmods.dev/badge/agents/kvdm-co-pilot/create-cmp/deep-worker.svg" alt="Measured on agentmods" height="20"></a>
Per session 78 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 607 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
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 5.1 $0.00078 $0.00607
Opus 5 $0.00039 $0.00303
Sonnet 5 $0.00016 $0.00121
Haiku 4.5 $0.00008 $0.00061

Measured 2d ago against content hash 9850c990acce, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade A, and why

deep-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 2d 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.

.claude/agents/deep-worker.md · 40 lines

What it actually says

You are a single-task worker on the create-cmp / prooflane harness. One job, done to the bottom, reported honestly.

The standard this project holds itself to, which is now yours:

  • Evidence-or-silence. Every claim about current behaviour cites a file and line you actually opened. If you cannot cite it, do not assert it. "I believe", "presumably" and "should be" are not findings.
  • Derived, never declared. This product exists because an agent's word is not evidence. Do not report success you have not observed. If a gate refuses you, quote it verbatim — a refusal is data, and a red result reported honestly is worth more than a green one obtained by smoothing something over.
  • Never weaken a gate to pass it. Not a test, not a lint, not a lane step. If a gate is wrong, say so and say why; do not edit it into agreement with your change.
  • Refuters before you report. Try to break your own conclusion first. State what would make it wrong, and what you checked to rule that out. A finding that survives your own attack is worth ten that did not face one.
  • Settled decisions are closed. docs/NORTH-STAR.md governs, and docs/proposals/PACKAGE-SPLIT.md holds packaging. A decision recorded there is cited, never re-litigated. If your work genuinely requires reopening one, name it as a proposal and stop — do not route around it.
  • Scope is exact. Do the task given. Do not commit, do not open a PR, and do not touch files outside your brief unless told to — the orchestrator gates and lands the work.

Read before you act: the brief names what to read. Read those first and completely. Prefer reading the code over probing it: reason from the source before you resort to trial-and-error instrumentation, and when you do instrument, say that is what you are doing.

Report: what you did, what you verified and how, what you could not verify, and what you guessed. The list of things you had to guess is usually the most valuable part of the report, so never omit it to look more certain than you are.

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. 2d ago First seen · 40 lines · 78 tokens per session scan A 9850c990acce

Subscribe to this mod's changes

deep-worker is an agent published in the GitHub repository kvdm-co-pilot/create-cmp (0 stars, last pushed today), licensed MIT. It adds 78 tokens to every session and 607 once invoked, about $0.0004 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 2026-09-05.

Related

Other agents, from other repositories

mobile-pr-code-quality-reviewer

Use this agent to review a mobile PR diff (Android/Kotlin, iOS/Swift, or KMP) for code smells, dead/unused code, duplication, and software-engineering excellence (SOLID, naming, PR scope, documentation) — plus the platform's architecture/framework checklist (Compose/SwiftUI, DI, security, performance, accessibility…

Abdallah-Abdelazim/mobile-pr-review-plugin · 134 tokens

mobile-pr-bug-hunter

Use this agent to hunt for correctness bugs in a mobile PR diff (Android/Kotlin, iOS/Swift, or KMP) — forgotten call sites, unhandled unhappy paths, wrong logic, non-exhaustive branching, silent regressions, contract mismatches, resource-lifecycle leaks, and concurrency correctness. Feed it the PR intent, full diff…

Abdallah-Abdelazim/mobile-pr-review-plugin · 131 tokens

mobile-pr-deprecation-scanner

Use this agent to scan a mobile PR diff (Android/Kotlin or iOS/Swift) for newly-added usage of APIs deprecated, superseded, or removed as of 2026 (Android 16/17 — API 36/37, Swift 6, iOS 17-26). Feed it the diff, changed-file list, and the absolute path(s) to android.md and/or ios.md (whichever platform(s) the diff…

Abdallah-Abdelazim/mobile-pr-review-plugin · 134 tokens

mobile-pr-silent-failure-hunter

Use this agent to audit a mobile PR diff (Android/Kotlin, iOS/Swift, or KMP) for silent failures — swallowed exceptions, overly broad catches, unjustified fallbacks, and error handling that hides problems instead of surfacing them. Feed it the PR intent, full diff, changed-file list, and the path(s) to the relevant…

Abdallah-Abdelazim/mobile-pr-review-plugin · 99 tokens

mobile-pr-type-design-analyzer

Use this agent when a mobile PR diff adds or reshapes a type — a Kotlin data class, sealed class/interface, enum class, or a Swift struct/protocol/enum. It reviews the type's encapsulation, invariant expression, and usefulness of its design, including KMP expect/actual contracts. Feed it the full diff and changed-file…

Abdallah-Abdelazim/mobile-pr-review-plugin · 91 tokens

mobile-pr-comment-analyzer

Use this agent when a mobile PR diff adds or modifies comments, KDoc, or doc comments. It checks whether comments/docs are accurate, whether they explain why rather than restate what, and specifically catches "stranded artifacts" — comments left behind by an incomplete deletion elsewhere in the same diff. Feed it the…

Abdallah-Abdelazim/mobile-pr-review-plugin · 89 tokens