mushi-mcp

mushi-mcp is a cursor rule for Cursor from kensaurus/mushi-mushi. It costs 499 tokens per session, scanned A, original, MIT.

A set of safety rules for coding agents using Mushi, a service for collecting and investigating user bug reports. It defines which actions can be read automatically and which changes require your confirmation.

In plain words
What is it for?
Use it to control Mushi access, inspect reports safely, keep fix pull requests as drafts, and handle missing credentials or permission errors correctly.
Why use it?
It reduces the risk of changing report status, replying to users, dispatching fixes, or exposing API keys without permission. It also prevents using a project credential outside the project it belongs to.

Cursor rule for Cursor

Written for Cursor: a Cursor rule (.mdc).

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 rules/kensaurus/mushi-mushi/mushi-mcp
Clone the repo
git clone --depth 1 https://github.com/kensaurus/mushi-mushi

Made for: Cursor.

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 mushi-mcp

README.md
[![agentmods](https://agentmods.dev/badge/rules/kensaurus/mushi-mushi/mushi-mcp.svg)](https://agentmods.dev/rules/kensaurus/mushi-mushi/mushi-mcp)
Your own site
<a href="https://agentmods.dev/rules/kensaurus/mushi-mushi/mushi-mcp"><img src="https://agentmods.dev/badge/rules/kensaurus/mushi-mushi/mushi-mcp.svg" alt="Measured on agentmods" height="20"></a>
Per session 499 This file is loaded in full into every session.
When invoked 499 The same file — it is already loaded in full.
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 5.1 $0.00499 $0.00499
Opus 5 $0.00249 $0.00249
Sonnet 5 $0.00100 $0.00100
Haiku 4.5 $0.00050 $0.00050

Measured 6d ago against content hash ce9f1f02d5f8, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

mushi-mcp 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 6d 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.

packages/cursor-plugin/rules/mushi-mcp.mdc · 42 lines

How it starts

The opening of the file, as written. The whole thing — 42 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Mushi MCP Safety Rules

Scope discipline

  • Always start with read-only tools: list_projects, get_recent_reports, get_report_detail, get_report_evidence, get_project_context, triage_issue.
  • Never call write tools (dispatch_fix, close_report, reopen_report, set_report_status, reply_to_reporter, dispatch_test_run, start_skill_pipeline, add_byok_key, award_bonus_points, set_tier) without explicit user confirmation in the same turn.
  • dispatch_fix must always use autoReadyPr: false by default so the PR stays a draft. Only set autoReadyPr: true if the user explicitly requests it.

Credential safety

  • Never echo, paste, log, or commit the value of MUSHI_API_KEY.
  • Do not suggest the user paste their API key into chat.
  • If a key is needed and unavailable, ask the user to set MUSHI_API_KEY in their environment and restart Cursor.

Project scope

  • API-key credentials are bound to a single project. Do not assume you can access other projects unless list_projects returns them.
  • If a tool returns INSUFFICIENT_SCOPE or PROJECT_REQUIRED, explain the error to the user and do not retry.

Evidence handling

  • Screenshot URLs and console logs may contain PII. Do not reproduce them verbatim in chat unless the user asks. Summarize instead.
  • Network request logs may contain auth tokens. Strip or redact Authorization values before displaying.

Fix dispatch discipline

Before calling dispatch_fix:

  1. Confirm you have read the full evidence (get_report_evidence or triage_issue).
  2. Confirm the recommended_actions list from triage_issue includes a fix recommendation.
  3. Show the user a one-paragraph summary of the proposed fix scope.
  4. Wait for explicit "yes" / "go ahead" / equivalent confirmation.

Sentry vs Mushi

Mushi sees user-felt bugs reported through its SDK (rage-clicks, console errors, network failures flagged by real users). Sentry sees thrown exceptions. They are complementary — use both when both are configured.

Read the full file on GitHub · 42 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. 6d ago First seen · 42 lines · 499 tokens per session scan A ce9f1f02d5f8

Subscribe to this mod's changes

mushi-mcp is a cursor rule published in the GitHub repository kensaurus/mushi-mushi (3 stars, last pushed 9d ago), licensed MIT. It adds 499 tokens to every session, about $0.0025 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.