koshi-quality-coach

koshi-quality-coach is an agent for Claude Code from jsharma1105/Koshi. It costs 55 tokens per session (1,151 once invoked), scanned A, original, MIT.

A Koshi quality-coach agent for evaluating AI interactions and tracking their quality over time. It scores retrieval, efficiency, caching, response speed, and user feedback, then groups results into grades.

In plain words
What is it for?
Use it to register teams, score AI turns, review quality trends, render team dashboards, and get context-engineering recommendations after enough feedback has been collected.
Why use it?
It helps teams identify recurring problems such as irrelevant, incomplete, outdated, verbose, or slow answers and decide what configuration changes may help.

Agent for Claude Code

Written for Claude Code: installed under .claude/.

Part of the koshi plugin — 5 agents shipped together

Good fit Use it to register teams, score AI turns, review quality trends, render team dashboards, and get context-engineering recommendations after enough feedback has been collected.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/jsharma1105/koshi/koshi-quality-coach
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.

Clone the repo
git clone --depth 1 https://github.com/jsharma1105/Koshi

Made for: Claude Code.

Or install koshi, the plugin that ships this one along with the rest of its 5 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 koshi-quality-coach

README.md
[![agentmods](https://agentmods.dev/badge/agents/jsharma1105/koshi/koshi-quality-coach/github.svg)](https://agentmods.dev/agents/jsharma1105/koshi/koshi-quality-coach)
Your own site
<a href="https://agentmods.dev/agents/jsharma1105/koshi/koshi-quality-coach"><img src="https://agentmods.dev/badge/agents/jsharma1105/koshi/koshi-quality-coach/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.

agentmods 80×15 button for koshi-quality-coach

Your own site · 80×15
<a href="https://agentmods.dev/agents/jsharma1105/koshi/koshi-quality-coach"><img src="https://agentmods.dev/badge/agents/jsharma1105/koshi/koshi-quality-coach.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 55 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,151 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
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 5.1 $0.00055 $0.01151
Opus 5 $0.00028 $0.00575
Sonnet 5 $0.00011 $0.00230
Haiku 4.5 $0.00006 $0.00115

Measured 8d ago against content hash 11b0b7d64ac3, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade A, and why

koshi-quality-coach 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 8d 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/koshi-quality-coach.md · 67 lines

How it starts

The opening of the file, as written. The whole thing — 67 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the Koshi Quality Coach — the per-team quality scoring and feedback specialist. You register teams, score their AI turns, surface trends, and recommend concrete config adjustments. Retrieval, memory, and context compilation are out of scope; hand those off.

Domain Knowledge

The 5-dimension composite score (0–1)

Dimension What it measures
Retrieval Did we surface enough relevant chunks? (retrievedChunks vs. team's topK)
Efficiency Budget utilization — overshooting and undershooting are both penalized. The sweet spot is ~0.7–0.85.
Cache cacheRatio of input tokens (cached / total)
Latency Total wall-clock latency; logarithmic decay above 2 s
User Explicit 1–5 star rating + any issue flags

Composite = weighted average → mapped to grades:

  • A ≥ 0.85 · B ≥ 0.70 · C ≥ 0.55 · D ≥ 0.40 · F < 0.40

Feedback issue flags (comma-separated)

irrelevant · incomplete · hallucinated · verbose · terse · format · outdated · slow

Trend analysis prerequisites

koshi_analyze_feedback needs at least 3 scored turns before it returns actionable trends.

Typical suggested adjustments

  • Weakest = Retrieval → increase topK (e.g. 5 → 7) or widen index scope.
  • Weakest = Efficiency → shrink tokenBudget, drop unused team context, or summarize history.
  • Weakest = Cache → switch packer to CacheOptimized, freeze team context.
  • Weakest = Latency → lower topK, smaller chunks, or move retrieval off the critical path.
  • Weakest = User → look at the issue flags; irrelevant/hallucinated ⇒ retrieval; verbose/terse/format ⇒ system prompt.

Tools you own

Tool Purpose
koshi_register_team(teamId, name, description?, tokenBudget?, topK?, qualityTarget?, systemPrompt?, teamContext?) Create a team profile
koshi_score_turn(teamId, retrievedChunks?, memoriesRecalled?, budgetUtilization?, cacheRatio?, latencyMs?, userRating?, issues?) Score one AI interaction
koshi_team_dashboard(teamId) Render trend dashboard + recommendations
koshi_analyze_feedback(teamId) Analysis + concrete suggested adjustments
koshi_list_teams() List all teams + their avg quality

Read the full file on GitHub · 67 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. 8d ago First seen · 67 lines · 55 tokens per session scan A 11b0b7d64ac3

Subscribe to this mod's changes

koshi-quality-coach is an agent published in the GitHub repository jsharma1105/Koshi (2 stars, last pushed 18d ago), licensed MIT. It adds 55 tokens to every session and 1,151 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.

Related

Other agents, from other repositories

checker

Judges finished work against its validator and the real need, leaving nothing unchecked. Use when code or a deliverable needs independent verification before it ships. Never edits the work, never implements the fix.

ai-driven-dev/framework · 42 tokens

executor

Turns a dispatched task into working, validated code that fits the project. Use when an approved scope must become code. Never plans, never judges its own work.

ai-driven-dev/framework · 34 tokens

MAF Migration Agent

Use when migrating a .NET codebase to Microsoft Agent Framework (MAF) 1.3.0. Orchestrates the full migration using specialized skills for API lookup, plan generation, CS0618 detection, and fan-out validation. Handles NuGet package updates, namespaces, executors, sessions, workflows, streaming, events, and DevUI guards.

joslat/maf-doctor · 78 tokens

MAF Auditor Agent

Use when you need to audit a .NET codebase and generate a ready-to-execute MAF migration plan. Given a repository, this agent scans source files and packages, runs dotnet-inspect diff, cross-references the Obsolete API Registry, and outputs a complete migration-plan.md with a populated tracking table.

joslat/maf-doctor · 69 tokens

MAF Best-Practice Reviewer

Use when you need a best-practice / drift audit of a clean MAF 1.3.0 codebase. Distinct from @maf-auditor (which generates a migration plan). This reviewer assumes you are ALREADY on 1.3.0 and asks 'is this code idiomatic, secure, observable, and identity-safe?' Output: audit-report.md, NOT migration-plan.md.

joslat/maf-doctor · 89 tokens

MAF Incident Responder

Use when a deployed MAF 1.3.0 app has hit a runtime failure (exception trace, hung workflow, silent workflow exit, agent loop, cost spike, prompt-injection symptom). Walks back from the symptom to the MAF pattern responsible, cross-references the obsolete-API registry + constraints, and proposes the deterministic fix.…

joslat/maf-doctor · 101 tokens