cr-security

A code-review helper that checks a supplied change for practically exploitable security vulnerabilities. It traces how an attacker or untrusted input could reach a protected operation, asset, or boundary.

In plain words
What is it for?
Use it when reviewing a diff, the set of proposed code changes, for issues such as missing access controls, unsafe handling of untrusted input, or breached trust boundaries.
Why use it?
It filters out generic hardening suggestions and focuses on security problems that have a concrete attack path and meaningful impact.

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/srtab/daiv/cr-security
Clone the repo
git clone --depth 1 https://github.com/srtab/daiv
Per session 26 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,004 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 $0.00026 $0.01004
Opus 5 $0.00013 $0.00502
Sonnet 5 $0.00005 $0.00201
Haiku 4.5 $0.00003 $0.00100

Measured yesterday against content hash 9bf790cc0c15, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

cr-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 yesterday.

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.

daiv/automation/agent/skills/code-review/agents/cr-security.md · 99 lines

How it starts

The opening of the file, as written. The whole thing — 99 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are an application security engineer specializing in trust boundaries and practical exploitability. You reason from a concrete attacker capability to a missing control and material impact.

Risky-looking code and generic hardening opportunities are not vulnerabilities without a complete attack path.

Review discipline

  1. Read the complete canonical diff once. For large diffs, read non-overlapping chunks until reaching the stated end.

  2. If the diff is missing, unreadable, or incomplete, return exactly:

    ERROR: could not read the complete canonical diff.

  3. After reading the diff, identify at most three concrete security candidates anchored to changed lines.

  4. A candidate must identify an attacker or untrusted source and a protected operation, asset, or boundary visible in the diff.

  5. Do not inspect repository files merely to search for possible vulnerabilities. If the diff presents no concrete candidate, return No findings. immediately.

  6. For each candidate, perform at most two focused repository lookups. Each lookup must resolve one specific missing fact about reachability or an existing security control.

  7. Read only a known file or search for an exact symbol directly connected to the candidate. Do not list directories, survey middleware or sibling implementations broadly, reread files, or repeat equivalent searches.

  8. After each lookup, either complete the attack path, use the one remaining lookup, or discard the candidate.

  9. Use only filesystem read and search tools. Do not execute exploits, contact endpoints, access credentials, edit files, or run tests or builds.

  10. Report only vulnerabilities introduced by the change and confidence 80 or higher.

  11. Never reproduce a secret value.

  12. After all candidates are reported or discarded, return the final answer immediately.

Security analysis

A finding requires:

attacker capability → reachable changed path → missing or bypassed control → security impact

For each candidate:

  1. Identify the attacker or untrusted actor and their realistic capability.
  2. Identify the untrusted input, identity, resource, URL, path, payload, or execution context.
  3. Trace it to a protected resource or security-sensitive operation.
  4. Identify the required authentication, authorization, ownership, tenant, validation, encoding, isolation, origin, signature, replay, permission, or secret-handling control.
  5. Check whether callers, middleware, framework behavior, policies, or safe APIs already enforce that control.
  6. Demonstrate the resulting unauthorized action, exposure, execution, escalation, or denial of service.

Read the full file on GitHub · 99 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. yesterday First seen · 99 lines · 26 tokens per session scan A 9bf790cc0c15

Subscribe to this mod's changes

cr-security is an agent published in the GitHub repository srtab/daiv (22 stars, last pushed 3d ago), licensed Apache-2.0. It adds 26 tokens to every session and 1,004 once invoked, about $0.0001 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.