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 the harvest sensor surfaces a candidate, /mos:qualify-opportunity puts the decision where it belongs — with you. The card shows WHY the candidate qualified: the Q1–Q8 rubric verdicts plus machine-readiness components, so you learn the qualification tests while you decide. Rejection teaches as much as approval — a Skip becomes graph data the ranker learns from. Nothing files without your explicit verb.

​
Usage

/mos:qualify-opportunity
/mos:qualify-opportunity review
This command is typically triggered automatically when /mos:opportunities scan surfaces a candidate, or you can invoke it directly to work through any pending harvested candidates in the queue.

​
The Qualification Card

Each candidate surfaces as a structured card showing two layers of evidence:

Q1–Q8 Rubric Verdicts

The eight human qualification checks: friction, connection, surprise, timing, actor fit, definability, newness, and desirability. Pedagogy over verdict — you see the reasoning behind each check, not just a score.

Component Readiness (D-18)

Machine readiness from existing measured signals: critic gate, compression, portfolio, tail flag, evidence readiness, and follow-through readiness. A missing input renders as unknown — never a fabricated zero.
The card also shows the advisory HarvestIndex_v1 score (marked EXPERIMENTAL), which combines rubric and readiness signals into a single ranking hint.

​
The Three Decisions

1

Qualify+file — approve and advance

Confirms the node as a human-promoted opportunity, advances its lifecycle to qualified, and files a bank entry. This is the only path that moves a candidate into the Opportunity Bank. After qualification, the opportunity waits in the bank until you explicitly trigger Explore.
2

Skip — reject with a reason

Writes a REJECTED_BECAUSE edge to the local graph with the failed-check reason. The node stays proposed and never resurfaces in future scans. The ranker learns from why-not as much as from yes.Skip reasons use a closed enum: q1_no_friction | q3_within_cluster | q4_stale_window | q5_no_actor | q7_already_filed | q8_disqualified | off_topic
3

Defer — hold for later

Holds the candidate without writing a rejection edge. It remains in the pending queue for a future qualification session.

​
Additional Card Verbs

Beyond the three primary decisions, the card also offers:
VerbWhat it does
Ask BrainConsults the teaching graph using generic framework handles only — no candidate prose crosses the wire. Recommended rather than triggered when Brain is absent.
RephraseHands the candidate back for re-titling before you decide.
Suggest nextMoves to the next-ranked candidate without deciding on the current one.

​
The Explore Action

Qualify and Explore are separate, explicit actions. Qualifying an opportunity does not auto-fire deep research — that is a deliberate cost-control boundary. A qualified opportunity waits in the bank until you explicitly trigger its exploration.
After you qualify an opportunity, the Explore action fetches web sources, cites them, and files the research in your opportunity bank. This is the path that turns a qualified candidate into cited deep-research evidence. It is always initiated by you — never automatic.

​
Why Rejection is Data

A Skip is not a dead end. When you reject a candidate, the reason becomes a typed REJECTED_BECAUSE edge in the room graph. Future scans are smarter because of it — the ranker learns from the why-not signal, and rejected candidates never resurface. The more you qualify and skip, the more precisely the scanner targets opportunities that fit your room.

​
Engine-Unavailable Fallback (D-20)

If the harvest engine cannot run (probe failures, missing substrate), the card offers one extra verb: [LLM manual scan (high effort)]. On acceptance, the model reads room artifacts directly and scores candidates against the same Q1–Q8 rubric. This is never the default and never a silent substitution. Every manual result carries engine_mode: llm_manual_baseline in node properties, the report provenance, and the bank entry frontmatter. Manual results are excluded from calibration sets.

​
Example

# Work through pending qualification candidates
/mos:qualify-opportunity

# Explicit review mode
/mos:qualify-opportunity review

/mos:opportunities

Scan for grant candidates and manage the Opportunity Bank pipeline.

/mos:research

Pull fresh web evidence to strengthen a qualified opportunity before Explore.

/mos:graph

Query the qualification history — see REJECTED_BECAUSE edges and qualification trails.

/mos:find-connections

Surface cross-domain analogies that might reframe an opportunity’s fit.

Build docs developers (and LLMs) love

Get started for freeTalk to us