/mos:pipeline — Compose and Run a Framework Chain

When you want several methodologies chained instead of run one-by-one, /mos:pipeline executes a multi-step pipeline with the room as the connecting tissue. It Brain-derives the framework sequence, composes it into resolver-verified /mos: commands, and runs the autonomous steps unattended — stopping only where human judgment is genuinely required. The Week 7 pattern for serious venture work.

Usage

/mos:pipeline --from-problem-type <type>
/mos:pipeline --from-framework "<framework name>"
/mos:pipeline <named-pipeline>

Problem type tokens — pass any of:

Alias	Canonical token
`undefined`	`UDP`
`ill-defined`	`IDP`
`well-defined`	`WDP`

/mos:pipeline --from-problem-type ill-defined
/mos:pipeline --from-framework "Beautiful Question Framework"

Entry modes

# From problem type — Brain traverses FEEDS_INTO from the problem class
/mos:pipeline --from-problem-type ill-defined

# From a specific framework — picks up mid-chain
/mos:pipeline --from-framework "Beautiful Question Framework"

When no arguments are passed, Larry reads room/STATE.md and recommends:

Discovery pipeline for Pre-Opportunity or Discovery stage
Thesis pipeline for Design or Investment stage

Larry presents both options and waits for your choice — no auto-selection.

What happens

Chain composition

The helper script calls recommendFrameworkChain (a FEEDS_INTO traversal — framework names and problem-type enums only, never command strings or user content), then composes the result into /mos: commands via composeWorkflow reading only the local data/command-registry.json. The resolved run order is printed before execution begins.

Resume check

Before Stage 1, reconcileResume(roomDir) reads room/.mindrian/pipeline-state.json — the sole chain-state source of truth. If a prior run exists, Larry offers: “I see you’ve already completed Stage N of the [chain] pipeline. Want to continue from Stage N+1? Or start fresh?”

Stage execution loop

For each stage: read the stage contract, extract structured input from the previous stage’s artifact, run the methodology command with that context pre-loaded, stamp the artifact with pipeline provenance metadata, then pause or continue per the autonomy gate.

Autonomy gate

Steps marked autonomous_safe: true run unattended. Non-safe steps pause with a “needs you here” gate — Larry explains why human judgment is required at this point before proceeding.

Pipeline complete

A summary of the full chain: what was produced at each stage, where each artifact was filed, and how each stage’s output fed the next as structured input.

The autonomy gate

The autonomy gate is Canon Part 3’s “human confirms” clause made literal. /mos:act --chain runs the autonomous_safe prefix unattended and stops at the first non-safe step. /mos:pipeline does the same inside its stage loop — every stop is a deliberate judgment checkpoint, not a limitation.

[PIPELINE] Stage 3 complete: JTBD → /mos:analyze-needs ✓

Stage 4: Minto Pyramid → /mos:minto
This step shapes the core investment thesis. Needs you here.

Bring forward from Stage 3:
  - 4 unmet jobs identified (IT integration, compliance overhead,
    reporting lag, cross-system reconciliation)
  - Primary persona: Operations Director

Ready to run Stage 4? [yes] [exit pipeline]

Command-less framework steps

If a framework in the chain has no corresponding /mos: command, the resolver prints:

no /mos: for <framework> — run it manually; continuing

The pipeline skips that step (or you run the framework manually) and continues to the next resolved command. It never invents a command for an unregistered framework.

Tier 0 behavior (no Brain)

When Brain MCP is not connected, the --from-problem-type and --from-framework helpers fall back to the local data/command-registry.json. The resolver still produces a valid command sequence — framework-only advice degrades gracefully through the same resolver path, never fabricating a command.

Standard chains

These chains are defined in the plugin’s pipeline architecture and available as named pipelines. Each arrow represents a FEEDS_INTO relationship where the left stage’s output becomes the right stage’s structured input.

Chain	Stages	Best for
A — Problem-to-Thesis	Domain Explorer → Bono → JTBD → Minto → DA → Investment	Well-defined problems ready for thesis framing
B — Discovery-to-Validation	HSI → Domain Explorer → Validation → Minto	Testing early hypotheses against market signals
C — System Analysis	Domain Explorer → Reverse Salient → JTBD → Bono → Minto	Wicked or ill-defined systemic problems
D — Full Venture	Domain Explorer → HSI → Bono → JTBD → Scenario → Minto → DA → Investment	Complete venture build from landscape to thesis

Pipeline chaining rules

The structured handoffs between stages — what each stage’s output becomes as the next stage’s input:

From	Output	To	Becomes
Domain Explorer	Sub-domains	Bono	Hats (one per perspective)
Domain Explorer	Sub-domains	JTBD	Personas (one per stakeholder)
Domain Explorer	Sub-domains	Scenario	Uncertainty axes
Bono	Hat insights	JTBD	Context per persona
Bono	Hat insights	Minto	Evidence for the pyramid
JTBD	Unmet jobs	Minto	SCQA complication
JTBD	Unmet jobs	Devil’s Advocate	Assumptions to test
Minto	Thesis	Devil’s Advocate	Target for destruction
Minto	Thesis	Investment Thesis	Core thesis
Devil’s Advocate	Survivors	Investment Thesis	Validated evidence
HSI	Bridges	Domain Explorer	New domains to map
Reverse Salient	Bottlenecks	Minto	Complication framing

Output / Artifacts

Each stage creates its own artifact — the pipeline is additive, never destructive. Artifacts are stamped with pipeline provenance in their YAML frontmatter:

pipeline: discovery
pipeline_stage: 3
pipeline_input: "4 unmet jobs from JTBD: IT integration, compliance overhead..."

The pipeline and pipeline_stage fields serve as the secondary confirming index for resume. The authoritative chain state lives in room/.mindrian/pipeline-state.json.

room/
  problem-definition/
    domain-explorer-YYYY-MM-DD.md     ← Stage 1 artifact
    jtbd-YYYY-MM-DD.md                ← Stage 3 artifact
  market-analysis/
    minto-YYYY-MM-DD.md               ← Stage 4 artifact
  .mindrian/
    pipeline-state.json               ← chain state (sole source of truth)

Pipelines are suggested sequences, not mandatory paths. You can exit at any stage — Larry summarizes what was completed and reminds you that resume is available. No guilt-trip about incomplete pipelines.

Example

/mos:pipeline --from-problem-type ill-defined

[PIPELINE] Brain-derived chain for problem type: ill-defined

Recommended chain (FEEDS_INTO traversal):
  Stage 1: Domain Explorer      → /mos:explore-domains
  Stage 2: Reverse Salient      → /mos:map-unknowns
  Stage 3: JTBD                 → /mos:analyze-needs
  Stage 4: Bono (Six Hats)      → /mos:bono
  Stage 5: Minto Pyramid        → /mos:minto

Autonomy map:
  Stages 1-2: autonomous_safe ✓ (will run unattended)
  Stage 3:    needs you here  ⚑
  Stages 4-5: autonomous_safe ✓

Start from Stage 1? [yes] [modify chain] [exit]

/mos:suggest-next

Get a ranked recommendation before committing to a full pipeline run.

/mos:grade

Score the room after the pipeline completes to measure what changed.

/mos:status

Check which sections are populated and healthy before choosing a chain.

/mos:graph

Inspect the FEEDS_INTO relationships that drive Brain-derived chain selection.

Overview

Room & Discovery

Problem Framing

Research & Intelligence

Grading & Strategy

/mos:pipeline — Compose and Run a Framework Chain

Usage

Entry modes

What happens

The autonomy gate

Command-less framework steps

Tier 0 behavior (no Brain)

Standard chains

Pipeline chaining rules

Output / Artifacts

Example

/mos:suggest-next

/mos:grade

/mos:status

/mos:graph

Build docs developers (and LLMs) love

Overview

Room & Discovery

Problem Framing

Research & Intelligence

Grading & Strategy

Documentation Index

​Usage

​Entry modes

​What happens

​The autonomy gate

​Command-less framework steps

​Tier 0 behavior (no Brain)

​Standard chains

​Pipeline chaining rules

​Output / Artifacts

​Example

​Related commands

/mos:suggest-next

/mos:grade

/mos:status

/mos:graph

Build docs developers (and LLMs) love

Usage

Entry modes

What happens

The autonomy gate

Command-less framework steps

Tier 0 behavior (no Brain)

Standard chains

Pipeline chaining rules

Output / Artifacts

Example

Related commands