Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/jsagir/mindrian-os-plugin/llms.txt

Use this file to discover all available pages before exploring further.

When you want a fast read on problem-discovery quality, /mos:grade scores the room on five components in under a minute. The lightweight cousin of a full calibrated assessment — it runs without Brain and delivers Larry’s honest evaluation of where your venture thinking stands right now.

Usage

/mos:grade
Add --full to dispatch eight parallel grading agents — one per room section — for a comprehensive venture-wide assessment with cross-section coherence scoring.
/mos:grade --full
autonomous_safe: true — this command runs without human-confirmation gates. Larry reads the room, scores it, and returns the result in one pass.

What happens

1

Calibration distribution preview

Before scoring, Larry shows a preview of the real grade distribution from 100+ past projects. Grades are not absolute — they’re relative to that distribution. You see where the calibration anchor sits before you see your score.
2

Room state analysis

Larry reads every populated section in the room silently — room/STATE.md plus all section artifacts. The entire venture is graded, not just one section.
3

Five-component scoring

Each component is scored 1–10, then weighted into a final grade. The scoring table is always shown — no exceptions.
4

Reality check

Every claim in the room is classified as Validated, Assumed, or Fantasy.
5

Grade computation

The weighted formula is applied and a letter grade with numeric score is produced.
6

Top 3 Actions

Specific next steps to improve the score, tied to the weakest component and the /mos: command that addresses it.

The five grading components

These are the Brain Grading Engine’s rubric weights, used for calibrated scoring against the real project corpus.
ComponentWeightWhat it measures
Technical Feasibility25%Can this actually be built or executed?
Logical Argument25%Does the reasoning hold together without leaps?
Tool Usage25%Were the right methodologies applied to the right problems?
Cognitive Bias12.5%Has the thinking been stress-tested against confirmation bias?
QA12.5%Are claims sourced, consistent, and internally coherent?
A low Tool Usage score (below 3/10) is the most common trigger for a Brain-recommended follow-up. If your score is low here, Larry will suggest the specific methodology that addresses it — typically Bono (De Bono Hats) or JTBD.

Brain mode vs. standard mode

Aspect/mos:grade (no Brain)/mos:grade (Brain connected)
Calibration dataNone — static rubricReal distribution from 100+ projects
Scoring model5-component staticDelegates to the Grading Agent
Percentile rankingNoYes
Assessment engineLarry directlyGrading Agent → Larry wraps results
When Brain MCP is active, Larry delegates to the Grading Agent which calls brain_grade_calibrate against the live project corpus and returns a percentile ranking alongside the component scores. Larry then wraps those results with teaching context and next steps.

Top feedback patterns

The Brain Grading Engine surfaces recurring patterns across real projects. If you see these in your feedback, they point to structural gaps rather than surface-level issues:
  • Vision-to-Execution Gap — strong hypothesis, no path to the first customer conversation
  • Framework Vomit — many methodologies applied but outputs not connected to each other
  • Solution-First — jumped to building before validating the problem worth solving
  • Single-Tool — relied on one methodology instead of a chained approach

Output / Artifacts

After scoring, Larry asks: “File this to problem-definition?” If confirmed, the grade artifact is written to room/**/grades/ with full provenance metadata including component scores, the reality-check classifications, and the top-3 actions.
room/
  problem-definition/
    grades/
      grade-YYYY-MM-DD.md   ← component scores, formula, top-3 actions

Example

/mos:grade
Larry reads all populated sections, shows the calibration distribution preview, then delivers the five-component scoring table:
[GRADE] Room: my-venture

Calibration anchor: median project scores 61/100 (C+)
Top decile: 84+ | Bottom decile: 43-

Component Scores
─────────────────────────────────────────────────
Technical Feasibility   7/10  (25%)  →  17.5 pts
Logical Argument        6/10  (25%)  →  15.0 pts
Tool Usage              3/10  (25%)  →   7.5 pts
Cognitive Bias          5/10 (12.5%) →   6.25 pts
QA                      6/10 (12.5%) →   7.5 pts
─────────────────────────────────────────────────
Total: 53.75 / 100   →  D+

Weakest component: Tool Usage (3/10)

Top 3 Actions
1. Run /mos:analyze-needs — only one methodology applied; chain it with JTBD to get real personas from your sub-domains.
2. Stress-test the market assumption — classified as Assumed, not Validated.
3. Add a Minto pyramid to tighten the logical argument structure.

/mos:suggest-next

Get Larry’s ranked recommendations for what to work on after seeing your grade.

/mos:pipeline

Run a multi-step methodology chain to systematically address the gaps your grade reveals.

/mos:status

Fast room health check — section coverage, health glyphs, and governing thoughts at a glance.

/mos:find-connections

Explore cross-section coherence to strengthen the logical argument component score.

Build docs developers (and LLMs) love