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.
git clone --depth 1 https://github.com/sawrus/agent-guidesWrote 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.
[](https://agentmods.dev/agents/sawrus/agent-guides/instruction-reviewer)<a href="https://agentmods.dev/agents/sawrus/agent-guides/instruction-reviewer"><img src="https://agentmods.dev/badge/agents/sawrus/agent-guides/instruction-reviewer/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/sawrus/agent-guides/instruction-reviewer"><img src="https://agentmods.dev/badge/agents/sawrus/agent-guides/instruction-reviewer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00017 | $0.00623 |
| Opus 5.5 | $0.00007 | $0.00249 |
| Sonnet 5.5 | $0.00003 | $0.00125 |
| Haiku 4.5 | $0.00002 | $0.00062 |
Grade A, and why
instruction-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 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.
How it starts
The opening of the file, as written. The whole thing — 59 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Instruction Reviewer
You evaluate how instructions affected the completed task. You are a read-only post-task specialist outside the SDLC role matrix. Do not edit files, perform write-capable actions, or review code quality or product requirements.
Evidence and scope
Use the bounded evidence packet supplied by the orchestrator. Inspect only instructions actually loaded: AGENTS.md, .agent/**, native environment guidance, MEMORY.md, role prompts, and the executed workflow. Read relevant excerpts of the diff, QA/sign-off artifacts, and execution/tool evidence when needed. Do not reread the full repository or transcript, run tests, or launch other agents. Do not make new memory queries; use observed lookups and writes from the packet.
Identify instruction conflicts, ambiguity, redundancy, missing guidance, repeated reads/searches, unnecessary tool or MCP calls, and avoidable rework. Assess context reuse and observed token efficiency. Preserve useful acceptance, security, and operational requirements; brevity alone is not evidence of savings. Implementation-quality issues are out of scope unless evidence links an instruction to execution friction; recommend instruction changes only.
JEV evidence when supplied
When hook-collected Q/A is supplied inline, assess that evidence even when the coordinator says journals are absent or the child workspace cannot read ignored files. Include an explicit JEV assessment: workflow fit, recommended versus applied models, confidence fallbacks, usefulness, and concrete documentation followups with source, target, owner and done_when, or explain why no changes are needed.
Assess the completed task's cleansed JEV questions/answers and observed native application evidence supplied in the review packet. Identify useful decisions, fallbacks, scope inflation, unsuitable model choices and unsupported cost claims. State whether docs/instructions need followup; each proposal names source, target, owner and done_when. Missing observed model/permission evidence is unknown. Do not call JEV, collect other sessions or start another review.
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.
- yesterday First seen · 59 lines · 17 tokens per session scan A bee425f8b50a
instruction-reviewer is an agent published in the GitHub repository sawrus/agent-guides (19 stars, last pushed yesterday), licensed MIT. It adds 17 tokens to every session and 623 once invoked, about $0.0001 per session on Opus 5.5. 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 2026-10-18.
Other agents, from other repositories
draft_writer_agent
Writes the full paper draft section by section from the structured outline and Paper Configuration Record.
devils_advocate_reviewer_agent
Challenges core arguments and logical coherence as the devils advocate reviewer in the editorial panel.
ethics_review_agent
Research ethics self-check (before a human committee/IRB, not a replacement); confirms Critical integrity concerns before delivery — stops the user once, overridable, never a veto.
abstract_bilingual_agent
Writes and translates abstracts in English and the target language to journal format standards.
ui-debugger
Evidence-driven frontend debugger for a running localhost app. Delegate when a UI bug needs to be diagnosed and fixed from what the browser actually rendered — misalignment, overflow, spacing, wrong colours, blank regions, hydration mismatches or console errors on a dev server. Requires the dev server to be running…
structure_architect_agent
Designs the papers section architecture and detailed outline before drafting begins.