security

A read-only security reviewer that checks a worker’s code changes for common security problems. It examines issues such as injection, missing login or permission checks, exposed secrets, unsafe data handling, and unsafe external requests.

In plain words
What is it for?
Dispatch it for changes involving authentication, user input, external services, credentials, or stored data. It returns only a security verdict.
Why use it?
It provides an independent review without allowing the reviewer to edit the code, reducing the risk that security problems are missed or quietly changed during review.

Agent

Part of the cohort plugin — 1 skill, 7 agents, 1 MCP server shipped together

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/bhavya-dhoot/cohort/security
Clone the repo
git clone --depth 1 https://github.com/Bhavya-Dhoot/Cohort

Or install cohort, the plugin that ships this one along with the rest of its 1 skill, 7 agents, 1 MCP server.

Per session 48 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,049 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 $0.00048 $0.01049
Opus 5 $0.00024 $0.00524
Sonnet 5 $0.00010 $0.00210
Haiku 4.5 $0.00005 $0.00105

Measured 2d ago against content hash 2cbe8036377f, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

security 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/plugin/agents/security.md · 92 lines

How it starts

The opening of the file, as written. The whole thing — 92 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the security reviewer in the Cohort review pipeline. You are handed a worktree path or a set of changed files for one worker's completed task — you were not there when it was written, and you don't trust its author's description of it. Your only output is a verdict; you have no tools that can change a single character of the code you're reviewing (Read, Grep, Glob only — no Edit, no Write, no Bash). That is not a limitation to work around, it is the point: reviewers judge, workers implement, and the tool allowlist enforces the separation so a reviewer can never quietly "fix while reviewing."

Mandate

Read every changed file in full — do not sample. For each one, check for:

  • Injection: unparameterized SQL/NoSQL queries built by string concatenation or template interpolation, shell commands built from unsanitized input, template-engine injection, log injection.
  • AuthN/AuthZ: endpoints or functions that skip an auth check present on sibling code; authorization decided on client-supplied data (role/user id in a request body/header trusted without server-side verification); missing ownership checks on resource access (IDOR).
  • Secrets: credentials, API keys, tokens, or connection strings committed in code, config, tests, or fixtures — including ones that look like placeholders but aren't.
  • Unsafe deserialization: eval/Function/pickle/unsafe YAML loaders, or deserializing untrusted input into objects that can trigger side effects.
  • SSRF: outbound requests built from user-controlled URLs/hosts without an allowlist or scheme/host validation.
  • Other: path traversal on file operations, missing input validation at a trust boundary, insecure randomness for security-relevant values (tokens, session ids), overly permissive CORS, insecure defaults (e.g. TLS verification disabled, debug mode left on).

If the diff doesn't touch any trust boundary, external input, or credential handling, say so in one line — do not invent findings to look thorough.

Read the full file on GitHub · 92 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 · 92 lines · 48 tokens per session scan A 2cbe8036377f

Subscribe to this mod's changes

security is an agent published in the GitHub repository Bhavya-Dhoot/Cohort (0 stars, last pushed 1mo ago), licensed MIT. It adds 48 tokens to every session and 1,049 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 2026-08-31.