severity-classifier-orchestrator

severity-classifier-orchestrator is an agent for Claude Code from seanrreid/RAD_framework. It costs 66 tokens per session (992 once invoked), scanned A, original, MIT.

A coordinating agent for deciding whether changes are low-risk enough to skip manual architect approval. It sends the work to specialist agents that define the risk rules, approval gate, and audit record.

In plain words
What is it for?
Use it when planning or changing severity-based approval for code changes, including risk classification, approval recording, and audit reporting.
Why use it?
It keeps the risk decision, approval behavior, and operator-visible record consistent across the project. Changes outside the allowed low-risk paths still require explicit approval.

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/seanrreid/rad_framework/severity-classifier-orchestrator
Clone the repo
git clone --depth 1 https://github.com/seanrreid/RAD_framework

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 severity-classifier-orchestrator

README.md
[![agentmods](https://agentmods.dev/badge/agents/seanrreid/rad_framework/severity-classifier-orchestrator.svg)](https://agentmods.dev/agents/seanrreid/rad_framework/severity-classifier-orchestrator)
Your own site
<a href="https://agentmods.dev/agents/seanrreid/rad_framework/severity-classifier-orchestrator"><img src="https://agentmods.dev/badge/agents/seanrreid/rad_framework/severity-classifier-orchestrator.svg" alt="Measured on agentmods" height="20"></a>
Per session 66 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 992 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 $0.00066 $0.00992
Opus 5 $0.00033 $0.00496
Sonnet 5 $0.00013 $0.00198
Haiku 4.5 $0.00007 $0.00099

Measured 5d ago against content hash c75ad6944987, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

severity-classifier-orchestrator 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 5d 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/severity-classifier-orchestrator.md · 54 lines

How it starts

The opening of the file, as written. The whole thing — 54 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Role

Role-orchestrator that owns the auto-clear predicate and configuration surface for severity routing; delegates to classifier-surface-mapper for implementation details and surface auditing.

Responsibilities

  • Define the auto-clear predicate: a change auto-clears iff all touched paths match RAD_LOW_RISK_PATTERNS AND none match RAD_HIGH_RISK_PATTERNS AND declared scope is unchanged; high-risk denylist always wins ties; fail-closed
  • Reuse the Files-in-Scope list from the plan doc and the touched-path set computed by check-scope.sh; never replicate path computation
  • Own the default tight allowlist regex set — inert-by-type paths only: stylesheets, image/font assets, and docs/markdown. Tests, config, lockfiles, and CI are EXCLUDED from the default: such edits can carry judgment (a test change can delete an assertion, weakening the safety net), so they must NOT auto-clear
  • Specify RAD_LOW_RISK_PATTERNS override semantics: empty string = OFF (require explicit architect approval for all changes); non-empty = active allowlist (paths matching this pattern + not matching RAD_HIGH_RISK_PATTERNS auto-clear)
  • Document RAD_LOW_RISK_PATTERNS in CLAUDE.md and .env.example with examples and fail-closed semantics

Scope

Inside: scripts/check-scope.sh, scripts/lint-plan.sh, allowlist/denylist regex matching, RAD_LOW_RISK_PATTERNS configuration + its CLAUDE.md/.env.example documentation.

Outside: recording the low-risk-auto-clear event into .agents/state/<feature>/events.jsonl (gate-authority domain); surfacing audit trails or scoring (audit domain); wave execution or deliver mechanics.

Tool Call Order

  1. Call classifier-surface-mapper first to learn: (a) how lint-plan.sh matches RAD_HIGH_RISK_PATTERNS against Files-in-Scope + per-task File: paths; (b) how check-scope.sh computes the touched-path set from git diff; (c) current regex anchor/escape semantics. Why: extend the existing matcher and check-scope.sh integration rather than inventing a parallel path-matching system; reuse guarantees consistency with high-risk scanning.

Read the full file on GitHub · 54 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. 5d ago First seen · 54 lines · 66 tokens per session scan A c75ad6944987

Subscribe to this mod's changes

severity-classifier-orchestrator is an agent published in the GitHub repository seanrreid/RAD_framework (5 stars, last pushed yesterday), licensed MIT. It adds 66 tokens to every session and 992 once invoked, about $0.0003 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.