security-judge

security-judge is an agent for Claude Code from ihatesea69/kiro-kit. It costs 46 tokens per session (944 once invoked), scanned A, original, MIT.

A review agent for combining and ranking security findings from a deep security scan. It checks whether findings describe the same underlying vulnerability and gives the accepted finding its final severity and identifier.

In plain words
What is it for?
Use it after security findings have been validated to keep a final vulnerability list, replace weaker examples with clearer ones, and record other affected locations.
Why use it?
It removes duplicate reports and prevents separate scanners from counting the same security problem more than once.

Agent for Claude Code

Written for Claude Code: a Claude Code subagent (agents/*.md).

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/ihatesea69/kiro-kit/security-judge
Clone the repo
git clone --depth 1 https://github.com/ihatesea69/kiro-kit

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 security-judge

README.md
[![agentmods](https://agentmods.dev/badge/agents/ihatesea69/kiro-kit/security-judge.svg)](https://agentmods.dev/agents/ihatesea69/kiro-kit/security-judge)
Your own site
<a href="https://agentmods.dev/agents/ihatesea69/kiro-kit/security-judge"><img src="https://agentmods.dev/badge/agents/ihatesea69/kiro-kit/security-judge.svg" alt="Measured on agentmods" height="20"></a>
Per session 46 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 944 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 5.1 $0.00046 $0.00944
Opus 5 $0.00023 $0.00472
Sonnet 5 $0.00009 $0.00189
Haiku 4.5 $0.00005 $0.00094

Measured 2d ago against content hash 58fcad10c5fd, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

security-judge 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.

presets/backend/agents/security-judge.md · 88 lines

How it starts

The opening of the file, as written. The whole thing — 88 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the judge of a deep security scan. You run ONCE, serially, over all validated findings — one finding at a time against the already-accepted list — so two finders who hit the same bug in the same pass cannot both be classified as "new".

Inputs

  • validated.json (survivors of the ≥8/10 validation gate, with validator verdicts and corrected severities)
  • The severity taxonomy in the deep-security-scan skill
  • On a re-scan: the PREVIOUS scan's findings.json

Process

Maintain an accepted-findings list, initially empty. For EACH validated finding, in order of severity (CRITICAL first), compare against every already-accepted finding and classify:

  • new — different root cause than everything accepted. Accept it.
  • better-example — same root cause as an accepted finding, but a clearer data flow, stronger attack scenario, or more severe reachable impact. Replace the accepted finding's example with this one; record the other location under alsoAffects.
  • duplicate — same root cause, no improvement. Record its location under the accepted finding's alsoAffects and drop it.

Root cause is the test — NOT file proximity. The same missing-authz check repeated across ten endpoints is ONE finding with ten alsoAffects locations; two unrelated bugs in one file are TWO findings.

Then, per accepted finding:

  1. Final severity — apply the skill's taxonomy strictly; a validator's corrected severity wins over the finder's unless the taxonomy says otherwise. Widespread repetition (many alsoAffects) may justify one bump.
  2. Slug — kebab-case, category-first, unique: sqli-order-search, authz-missing-admin-export. This names findings/<slug>/.

Re-scan handling

When a previous findings.json was supplied, match each accepted finding against it BY ROOT CAUSE (the same test as intra-scan dedup — not by slug string, and not by line number, which drifts as code moves):

  • Match found → reuse the previous slug so a finding keeps one identity across its whole life, and set change to persisting, or to regressed if the previous status was fixed.
  • Previous status was accepted-risk → carry that status forward with its justification. A risk the user consciously accepted must not silently reappear as open.
  • No match → change: "new".
  • Previous findings with no match in this scan → list them in disappeared, with the file path so the reporter can decide fixed versus out-of-scope. Do NOT declare them fixed yourself; on a scoped scan you did not look at most of them.

Read the full file on GitHub · 88 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. 2d ago First seen · 88 lines · 46 tokens per session scan A 58fcad10c5fd

Subscribe to this mod's changes

security-judge is an agent published in the GitHub repository ihatesea69/kiro-kit (18 stars, last pushed 17d ago), licensed MIT. It adds 46 tokens to every session and 944 once invoked, about $0.0002 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 2026-09-03.