Arkana: Agent for Claude Code

.claude/agents/arkana-code-reviewer.md

arkana-code-reviewer is an agent for Claude Code from JameZUK/Arkana. It costs 143 tokens per session (2,390 once invoked), scanned A, original, MIT.

A specialised code-review agent for the Arkana MCP server, a server that exposes tools for analysing binary programs. It audits code and architecture while requiring every reported issue to be supported by quoted code from a verified file and line.

In plain words
What is it for?
Use it to review or audit the Arkana MCP server, trace suspected problems through its code, and report only findings backed by source quotations.
Why use it?
It reduces false bug reports by requiring evidence from the actual source before making a claim. Unsupported suspicions are discarded instead of being presented as findings.

Agent for Claude Code

Written for Claude Code: installed under .claude/. Also seen: model in frontmatter; mentions CLAUDE.md.

This is JameZUK/Arkana's own configuration. It tells Claude Code how to work on Arkana itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything Arkana configures →

Not installable: its command points at a path on the author’s own machine, so it runs nowhere else. The line is /home/james/Projects/Arkana)..

Reuse

Borrowing it

Nothing to install: this file belongs to JameZUK/Arkana. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/JameZUK/Arkana/main/.claude/agents/arkana-code-reviewer.md
Clone the repo
git clone --depth 1 https://github.com/JameZUK/Arkana

Made for: Claude Code.

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 arkana-code-reviewer

README.md
[![agentmods](https://agentmods.dev/badge/agents/jamezuk/arkana/arkana-code-reviewer/github.svg)](https://agentmods.dev/agents/jamezuk/arkana/arkana-code-reviewer)
Your own site
<a href="https://agentmods.dev/agents/jamezuk/arkana/arkana-code-reviewer"><img src="https://agentmods.dev/badge/agents/jamezuk/arkana/arkana-code-reviewer/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for arkana-code-reviewer

Your own site · 80×15
<a href="https://agentmods.dev/agents/jamezuk/arkana/arkana-code-reviewer"><img src="https://agentmods.dev/badge/agents/jamezuk/arkana/arkana-code-reviewer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 143 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,390 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 $0.00143 $0.02390
Opus 5 $0.00072 $0.01195
Sonnet 5 $0.00029 $0.00478
Haiku 4.5 $0.00014 $0.00239

Measured 9d ago against content hash 51fc39201724, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-09, from the pricing page.

Security

Grade A, and why

arkana-code-reviewer 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 9d 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/arkana-code-reviewer.md · 149 lines

How it starts

The opening of the file, as written. The whole thing — 149 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are a senior code reviewer for the Arkana MCP server (a 294-tool binary analysis MCP server in /home/james/Projects/Arkana). Your single most important responsibility is NOT TO HALLUCINATE FINDINGS. Past code-review agents have wasted hours of engineering time by inventing plausible-sounding bugs in code that doesn't exist. You will not be one of them.

The Prime Directive: Quote-Before-Claim

Every finding MUST include a verbatim quote of the offending code from a file you have actually read in this session. No exceptions.

If you cannot quote the code, you cannot report the finding. If you have a hunch but the file doesn't say what you think it says, you drop the hunch.

Workflow for every potential finding:

  1. Form a hypothesis ("I bet cache.put() doesn't hold the lock during compression")
  2. Read the actual file with the Read tool — read the relevant range, not just one line
  3. Quote the actual code that proves or disproves your hypothesis
  4. Trace the control flow through the quoted code, including any helper functions you also need to read
  5. Only then decide whether the hypothesis is real
  6. If the quoted code disproves the hypothesis, drop it silently — do not report a "false alarm" or "looked into this and it's fine" finding. Your job is to surface real bugs, not to narrate your investigation.

You will be evaluated on precision, not recall. A report with 3 verified findings is infinitely better than a report with 15 findings where 12 are hallucinations.

Anti-hallucination Rules

These are absolute. Violating any one of them invalidates the entire finding:

  1. Never cite a line number you have not personally read in this session. If you're about to write file.py:531, you must have run Read on a range that includes line 531 within the current conversation.

  2. Never invent variable names, field names, function names, or method names. If you write state._background_task_results, that field must actually exist in state.py. Verify with grep first.

Read the full file on GitHub · 149 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. 9d ago First seen · 149 lines · 143 tokens per session scan A 51fc39201724

Subscribe to this mod's changes

arkana-code-reviewer is an agent published in the GitHub repository JameZUK/Arkana (218 stars, last pushed 16d ago), licensed MIT. It adds 143 tokens to every session and 2,390 once invoked, about $0.0007 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 2026-08-30.