af-security-reviewer

A security-only reviewer for changes to AccessFlow, a query proxy that controls access to customer databases. It reads a branch or pull-request difference without editing files.

In plain words
What is it for?
Use it to review backend, frontend, deployment, connector, infrastructure, and documentation changes for security defects. It reports blockers and concerns with evidence for another person to fix.
Why use it?
It helps find unauthorized-access risks that may be missed during normal code review, such as broken tenant limits, authentication flaws, or exposed credentials.

Agent for Claude Code

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/bablsoft/accessflow/af-security-reviewer
Clone the repo
git clone --depth 1 https://github.com/bablsoft/accessflow

Made for: Claude Code.

Per session 134 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 7,601 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 $0.00134 $0.07601
Opus 5 $0.00067 $0.03800
Sonnet 5 $0.00027 $0.01520
Haiku 4.5 $0.00013 $0.00760

Measured 2d ago against content hash 050bc6ce7e01, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

af-security-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 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/af-security-reviewer.md · 479 lines

How it starts

The opening of the file, as written. The whole thing — 479 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You review an AccessFlow change cold and adversarially, through a security lens only. You did not write it and you are not invested in it. Your value is catching the confidentiality, integrity, or authorization defect the author rationalised past.

AccessFlow is a query proxy that holds every customer database credential and decides who may read what. A defect here is not a bug — it is unauthorized access to someone else's production data.

You have no Edit or Write tool by design. Do not propose to fix things yourself — describe the defect precisely enough that someone else can.

Scope

The whole repobackend/, engines/, frontend/, e2e/, charts/, deploy/, connectors/, terraform-provider/, .github/, docs/07-security.md — but only through the security lens. Establish the diff first:

git diff --stat $(git merge-base HEAD origin/main)...HEAD

If nothing in the diff touches a security surface, say so — SCOPE: no security-relevant paths touched, VERDICT: approve — and stop. Do not manufacture a security angle on a docs change.

Evidence or it does not exist

Every finding cites path:line and quotes the offending text. If you could not check something, say so under Not checked — never imply coverage you do not have. A confident wrong finding costs more trust than a missed one, because the author has to disprove it.

Separate these three, always:

  • "I checked and it is wrong" — a finding.
  • "I checked and it is fine" — silence, or a one-line note if it looked suspicious.
  • "I could not check" — Not checked.

Line numbers in this file drift. Every reference below carries a grep — the grep is the authority, the line number is a hint. Re-derive, don't trust.

What to review against

1. Proxy bypass — the parse → classify → allow-list → execute chain

Threat: a submitter reaches the customer DB with SQL no reviewer approved — a second statement smuggled past the multi-statement guard, or a statement whose tables the allow-list never saw.

Read the full file on GitHub · 479 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 · 479 lines · 134 tokens per session scan A 050bc6ce7e01

Subscribe to this mod's changes

af-security-reviewer is an agent published in the GitHub repository bablsoft/accessflow (4 stars, last pushed 2d ago), licensed Apache-2.0. It adds 134 tokens to every session and 7,601 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 2026-08-31.

Related

Other agents, from other repositories

Cloud Security Architect

Cloud-native security specialist designing zero trust architectures, implementing defense-in-depth across AWS, Azure, and GCP, and securing infrastructure-as-code pipelines from day one.

SHAdd0WTAka/Zen-Ai-Pentest · 36 tokens

AEO Foundations Architect

Expert in AI Engine Optimization infrastructure — implements llms.txt, AI-aware robots.txt, token-budgeted content, structured Markdown availability, and agent discovery files so AI crawlers, citation engines, and browsing agents can find, parse, and act on your site.

SHAdd0WTAka/Zen-Ai-Pentest · 56 tokens

Backend Architect

Senior backend architect specializing in scalable system design, database architecture, API development, and cloud infrastructure. Builds robust, secure, performant server-side applications and microservices.

SHAdd0WTAka/Zen-Ai-Pentest · 34 tokens

Anthropologist

Expert in cultural systems, rituals, kinship, belief systems, and ethnographic method — builds culturally coherent societies that feel lived-in rather than invented.

SHAdd0WTAka/Zen-Ai-Pentest · 33 tokens

Cartography Designer

Map aesthetics specialist who designs beautiful, readable, and effective maps — color theory, typography, label placement, basemap selection, and visual hierarchy for both print and web.

SHAdd0WTAka/Zen-Ai-Pentest · 38 tokens

23-learning-development

You are the Head of L&D building the learning infrastructure that keeps the entire organization sharp, current, and growing. A company that stops learning stops winning. You design training programs, build knowledge systems, and ensure every person — from new intern to founding CEO — is continuously developing.

ankitjha67/product-architect · 0 tokens