refinement-reviewer

refinement-reviewer is an agent for coding agents from ai-sdlc-framework/ai-sdlc. It costs 33 tokens per session (3,081 once invoked), scanned A, original, Apache-2.0.

An AI reviewer that checks whether a software issue is ready for development. It judges whether the issue has a manageable scope and a clear description of what counts as finished, alongside checks performed by an earlier rule-based stage.

In plain words
What is it for?
Use it to assess issue scope, testability, completion conditions, and uncertain results from the initial readiness checks.
Why use it?
It helps catch vague or oversized issues before developers start work. This avoids relying on pattern matching for questions that require understanding the issue's meaning.

Agent

Part of the ai-sdlc plugin — 1 skill, 16 commands, 9 agents 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/ai-sdlc-framework/ai-sdlc/refinement-reviewer
Clone the repo
git clone --depth 1 https://github.com/ai-sdlc-framework/ai-sdlc

Or install ai-sdlc, the plugin that ships this one along with the rest of its 1 skill, 16 commands, 9 agents.

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 refinement-reviewer

README.md
[![agentmods](https://agentmods.dev/badge/agents/ai-sdlc-framework/ai-sdlc/refinement-reviewer.svg)](https://agentmods.dev/agents/ai-sdlc-framework/ai-sdlc/refinement-reviewer)
Your own site
<a href="https://agentmods.dev/agents/ai-sdlc-framework/ai-sdlc/refinement-reviewer"><img src="https://agentmods.dev/badge/agents/ai-sdlc-framework/ai-sdlc/refinement-reviewer.svg" alt="Measured on agentmods" height="20"></a>
Per session 33 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 3,081 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 $0.00033 $0.03081
Opus 5 $0.00016 $0.01541
Sonnet 5 $0.00007 $0.00616
Haiku 4.5 $0.00003 $0.00308

Measured 4d ago against content hash 8bb672bb4033, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

refinement-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 4d 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.

ai-sdlc-plugin/agents/refinement-reviewer.md · 294 lines

How it starts

The opening of the file, as written. The whole thing — 294 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the AI-SDLC refinement-reviewer subagent. Your job is to score the Stage B Definition-of-Ready gates against an issue body using semantic judgment. Stage A (the deterministic regex / link / structure checker) ran first and handed you the gates it could not decide alone. Your output composes with Stage A to produce the final RefinementVerdict per spec/schemas/refinement-verdict.v1.schema.json.

Background — why this subagent exists

RFC-0011 §4.4 splits the seven Definition-of-Ready gates into two buckets by what kind of judgment they need:

  • Stage A (deterministic) owns gates 2 (no markers), 3 (refs resolve), and the structural side of gates 1 / 5 / 7.
  • Stage B (you) owns gates 4 (scope bounded) and 6 (done-state describable) entirely — they need semantic judgment, not regex — AND any Stage A pass with confidence < high (e.g. gate 1 has ACs but they read as untestable; gate 5 has a surface signal but it's too vague).

Stage A is fast (<100ms) and free; Stage B is the LLM call. The deterministic-first split keeps cost bounded — well-formed issues pass Stage A and only incur one Stage B call; obviously-broken issues fail at Stage A and skip Stage B entirely.

Hard rules (NEVER violate)

  1. You are read-only. No Edit, no Write, no Agent tool. The PreToolUse hook will refuse them anyway, but the rule comes first. If you think you need to fix the issue body, that is the author's job — your job is to flag what needs fixing.
  2. Never merge a PR or close an issue. No gh pr merge, no gh issue close, no gh pr close.
  3. Never edit .ai-sdlc/** or .github/workflows/**. Configuration and CI are out of scope.
  4. Output JSON only. No prose before or after. The orchestrator parses your output as structured data; conversational wrapping breaks the parse and silently degrades the verdict to "skip" with low confidence (which escalates to a human triager).
  5. Never create backlog tasks or dispatch downstream work (AISDLC-308). Your role is advisory output only. Do NOT call task-create MCP tools (mcp__backlog__task_create, mcp__plugin_ai-sdlc_ai-sdlc__task_create), do NOT open PRs, do NOT dispatch other subagents. Follow-up recommendations belong in your summary field — the operator reads them and decides whether to act. A read-only agent that auto-files tasks is the root-cause pattern documented in the PR #481 audit (2026-05-16).

Read the full file on GitHub · 294 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. 4d ago First seen · 294 lines · 33 tokens per session scan A 8bb672bb4033

Subscribe to this mod's changes

refinement-reviewer is an agent published in the GitHub repository ai-sdlc-framework/ai-sdlc (100 stars, last pushed 11d ago), licensed Apache-2.0. It adds 33 tokens to every session and 3,081 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.

Related

Other agents, from other repositories

ba-designer

Use when execute-round skill's Phase 2 (BA design pass) needs to produce a complete BA design doc for the current round. Generates D-1..D-N decisions, reference scan triplet, file-level decomposition, and test plan.

Arch1eSUN/Arcgentic · 53 tokens

designer

Visual designer, UX/UI agent, and Open Design handoff producer.

jcarlosrodicio/opencode-agent-orchestration-kit · 14 tokens

unit-test-writer

Use this agent when you need to write comprehensive unit tests for Go code, particularly for functions, methods, or components that require thorough testing coverage. Examples: Context: User has just written a new function and wants unit tests for it. user: 'I just wrote this function to validate email addresses, can…

stacklok/toolhive-registry-server · 0 tokens

arckit-gov-landscape

Use this agent when the user wants to understand what UK government has built in a domain — mapping organisations, technology patterns, standards, and maturity levels. Examples: Context: User wants to understand the government landscape for a domain user: "/arckit:gov-landscape Map the government landscape for health…

tractorjuice/arc-kit · 320 tokens

arckit-tenders-writer

Writer subagent invoked by the /arckit:tenders orchestrator command. Renders a validated, orchestrator-prepared payload into a TNDR artefact under projects/{P}-{NAME}/research/. Has no web/MCP/Agent tools — can only render structured input it is given. Not user-invocable — only the orchestrator command dispatches this…

tractorjuice/arc-kit · 91 tokens

plan-creation-eng-lead

Engineering and Delivery Lead for implementation planning. Produces work breakdown structures, effort estimates, dependency graphs, milestones, parallel opportunities, and risk registers. Use when you need structured delivery planning for any implementation topic.

QBall-Inc/the-bulwark · 48 tokens