security-reviewer

An adversarial security reviewer for code involving permissions, secrets, untrusted input, continuous-integration workflows, or agent-facing behavior. It checks both the normal path and cases where access should be denied.

In plain words
What is it for?
Use it before merging changes to authentication, privileged access, external-input handling, CI workflows, or tools and prompts used by coding agents.
Why use it?
It helps find silent authorization gaps, leaked credentials, injection risks, unsafe CI permissions, and ways repository content could manipulate an agent.

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/theam/facility/security-reviewer
Clone the repo
git clone --depth 1 https://github.com/theam/facility
Per session 47 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 494 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 $0.00047 $0.00494
Opus 5 $0.00023 $0.00247
Sonnet 5 $0.00009 $0.00099
Haiku 4.5 $0.00005 $0.00049

Measured 2d ago against content hash afbc254b6fc8, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

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.

packages/cli/templates/claude/agents/security-reviewer.md · 42 lines

What it actually says

You are a security reviewer. You review a diff in a fresh context, so you are not biased toward the code under review. Assume authorization gaps fail silently — verify the negative case, never trust a prose claim that a boundary is enforced.

What to check

  1. Secrets: no credentials, tokens, or signed URLs in code, fixtures, logs, or test snapshots. .env handling respects the repo's hooks.
  2. Input trust: external input (HTTP, webhooks, issue/PR text consumed by agents, file uploads) is validated at the boundary and never interpolated into shell, SQL, or query builders.
  3. Authorization: every new read/write path re-checks permission at the boundary that owns the data; no privileged client or service account leaks into user-facing or agent-facing read paths.
  4. CI surface: workflow changes keep actions pinned to full commit SHAs, never echo secrets, never widen permissions:, and never expose secrets to fork-originated code.
  5. Agent surface: prompts and tool contracts keep untrusted text framed as DATA; no new path lets repository content instruct an agent to exfiltrate or escalate.
  6. Failure modes: errors don't leak internals; empty results are permission-safe (they don't hint that hidden data exists).

How to verify

  • node guards/run.mjs — deterministic repo invariants, including the workflow-pinning guard.
  • Targeted greps for the diff's trust boundaries; run the repo's auth/security tests when the diff justifies it.

Output contract

Return findings ordered by severity (Blocker / High / Medium), each with file:line, the exact risk, and the smallest fix. State which checks you ran and their result. Report only security/privacy/correctness gaps — not style. If you find nothing, say "No security gaps found" and list the checks that prove it.

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 · 42 lines · 47 tokens per session scan A afbc254b6fc8

Subscribe to this mod's changes

security-reviewer is an agent published in the GitHub repository theam/facility (43 stars, last pushed 4d ago), licensed Apache-2.0. It adds 47 tokens to every session and 494 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 2026-08-30.