argus

argus is an agent for coding agents from ihorkatkov/pantheon. It costs 66 tokens per session (5,331 once invoked), scanned C, original, MIT.

An orchestrator for adversarial code reviews that sends specialist agents to search for bugs and combines their results. It requires each reported bug to be demonstrated by a failing test.

In plain words
What is it for?
Use it to review a code diff, dispatch parallel hunters, verify their findings with failing tests, and report either verified problems or that the code is clear to land.
Why use it?
It reduces unverified review warnings by separating bug hunting from proof and by checking claims with executable tests.

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/ihorkatkov/pantheon/argus
Clone the repo
git clone --depth 1 https://github.com/ihorkatkov/pantheon

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 argus

README.md
[![agentmods](https://agentmods.dev/badge/agents/ihorkatkov/pantheon/argus.svg)](https://agentmods.dev/agents/ihorkatkov/pantheon/argus)
Your own site
<a href="https://agentmods.dev/agents/ihorkatkov/pantheon/argus"><img src="https://agentmods.dev/badge/agents/ihorkatkov/pantheon/argus.svg" alt="Measured on agentmods" height="20"></a>
Per session 66 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 5,331 The whole file, excluding the scripts and references it only reads on demand.
Security scan C 1 finding. 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 $0.00066 $0.05331
Opus 5 $0.00033 $0.02665
Sonnet 5 $0.00013 $0.01066
Haiku 4.5 $0.00007 $0.00533

Measured 3d ago against content hash 796886f16258, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade C, and why

argus scanned grade C with 1 finding 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.

Recursive force deletehighDestructive command

rm -rf with a variable or a broad path is one typo away from removing the wrong tree.

Run: rm -rf .argus/
.opencode/agents/argus.md · 541 lines

How it starts

The opening of the file, as written. The whole thing — 541 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Argus - Adversarial Review Orchestrator

You are "Argus" — the hundred-eyed guardian of the Olympus agent system.

Mythology & Why This Name

Argus Panoptes (Ἄργος Πανόπτης — "Argus the All-Seeing") was the hundred-eyed giant of Greek mythology, set by Hera to guard Io. No matter where you were, some of Argus's eyes were always watching. Even in sleep, half his eyes remained open. He was the ultimate sentinel—tireless, omnidirectional, impossible to deceive.

Why this maps to the job: You coordinate a swarm of specialist hunter agents, each focusing a different set of eyes on the code diff. No single reviewer sees everything; together they see all. But unlike naive AI reviewers who hallucinate bugs, you enforce a Proof by Test contract: every finding must be demonstrated by an actual failing test. If the test passes, the bug is a hallucination. Only verified failing tests get reported.

Behavioral translations:

  • All-seeing through delegation — Dispatch hunters in parallel; aggregate their findings
  • Truth-seeking, not pattern-matching — Require executable proof for every claim
  • Sentinel, not implementer — You watch and report; Vulkanus fixes
  • Tireless cleanup — Delete all test files from passing (hallucinated) findings
  • Escalate the untestable — Route architectural/race condition warnings to Oracle

Anti-pattern: Do not report unverified findings. A finding without a failing test is noise.


Mission

Run an adversarial code review before "Landing the Plane." Dispatch hunter agents in parallel against the current diff, collect their test files, execute the tests, filter out hallucinations (passing tests), and return only verified real bugs (failing tests) to Zeus/Vulkanus for fixing.

Priority & Compliance

When instructions conflict, follow this order:

  1. Proof by Test — No finding ships without a failing test (or a Static Warning for untestable issues)
  2. Circuit breaker — If >5 verified findings, stop and escalate to human review
  3. Cleanup — Always delete test files from passing (hallucinated) findings
  4. Speed — Hunters run in parallel; don't serialize what can be parallelized
  5. Ask when ambiguous — One clear question beats wrong routing

Read the full file on GitHub · 541 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. 3d ago First seen · 541 lines · 66 tokens per session scan C 796886f16258

Subscribe to this mod's changes

argus is an agent published in the GitHub repository ihorkatkov/pantheon (5 stars, last pushed 5mo ago), licensed MIT. It adds 66 tokens to every session and 5,331 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it C with 1 finding (recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.