A&ID Diagrams

Dreamineering as a visible operating system

The diagrams page is the shop window for the doctrine: Stackmates is tech that plays for each other. Intention enters, the tight five shape committed work, and evolution exits as a sharper standard for the next agent.

Canonical doctrine belongs in the AISL and FLOW docs. This route sells and tests the visual language by assembling agency-owned metadata with ui-symbols renderers.

§01

System of dreamineering

The high-level place is the FLOW/AISL doctrine; this page makes it inspectable. INTENTION and EVOLUTION are rails, not engine phases. The tight five are the only executable loop.

Entry rail

INTENTION

Name the evolution sought before the engine starts. If the desired standard cannot be named, the idea is not ready for action.

Engine

The tight five

  1. 1. Capture: problems and potential.
  2. 2. Prioritize: what next, best value.
  3. 3. Commit-action: focus attention.
  4. 4. Measure outcome: results vs predictions.
  5. 5. Question-learn: better questions, better insights.

Exit rail

EVOLUTION

The run only counts when a standard, workflow, rule, skill, doc, or product constraint is now permanently sharper. That is how creative intent crosses into practical Stackmates work without becoming route-and-ship drift.

§02

Language key

The key stays close to the diagrams so the symbol language can evolve while the docs stay readable.

Agents — the doers

HAHuman Agent
DADigital Agent
PAPhysical Agent
OAOrchestrator Agent

Instruments — verify & reward

GGate
VValidator
MMonitor
RRecorder
CController
AAlarm
SSelector
TTransmitter

Connections

Value flow
Signal flow
Feedback flow

§03

Flow of intent

The operating loop that carries INTENTION through the tight five and exits as EVOLUTION: a permanent improvement to the system.

Flow of intent

operating-model

INTENTION to EVOLUTION, with the tight five as the engine: capture, prioritize, commit-action, measure outcome, question-learn.

Source: docs/apps/aisl-diagrams/2-dream.md

CAP-R01Capture
DAWorking agent[L2]●
capture

Problems and potential are preserved before the system tries to act.

shape
PRI-S01Prioritize
DAWorking agent[L2]●
route

Choose what next by best value: demand, cost, risk, reward, and timing.

exit decision
ACT-G01Commit action[FC]
DAWorking agent[L2]●
commit

Focus attention on one move; no artifact is written without naming its true exit home.

receipt
MSR-V01Measure outcome
DAWorking agent[L2]●
measure

Compare predicted result with the actual receipt.

question
LRN-C01Question learn
DAWorking agent[L2]●
learn

Ask the better question that would have made the run cheaper, clearer, or faster.

Feedback controls

measure-outcome
question
question-learn
question-learn
[next run]sharper capture
capture-ideas
  • INTENTION and EVOLUTION are rails, not engine phases.
  • The tight five are the only executable sequence.
  • If no standard, workflow, rule, or artifact changed, the run did not evolve.

§04

Agency triad

Three schematics document the current state, intended architecture, and bridge path in the same visual vocabulary.

Agency current reality

reality

The live library mixes algorithms, workcharts, orchestration runtime, protocols, prompt assets, and product-specific support behind a broad root barrel.

Source: docs/architecture/libs/agency/01-current-reality.md

Agents

HAMaintainer[L3]●

Reads the reality map before touching agency architecture.

OALegacy orchestrator[L2]○

Still present as transitional execution/control vocabulary.

DAMixed workchart services[L1]→→

Some execute real AI paths; others only prove scaffold flow.

Instruments

API-S01661-export root barrel

Selection surface is too broad; consumers import wholesale.

PRF-V01Intent proof[FO]

Mostly proves contract/scaffold flow, not AI value.

VAL-A01Value-proof alarm[FO]

Flags template outputs, scripted eval scores, and dormant consumers.

Flows

Maintainer
[scan]inventory
661-export root barrel
661-export root barrel
activation path
Mixed workchart services
Intent proof
[F]not value proof
Value-proof alarm
  • Reality is a measured map, not a target-state proposal.
  • The critical rot is mixed purpose and weak proof semantics.
  • Workchart movement must stop rewarding scaffold-only proofs.

Agency dream architecture

dream

Agency becomes a small intelligence kernel: algorithms decide, workcharts assemble, pipelines execute, protocols govern, middleware proves, application surfaces activate.

Source: docs/architecture/libs/agency/02-dream-architecture.md

Agents

DAApplication/API/CLI consumer[L1]●

Calls narrow supported agency subpaths instead of the root barrel.

OAWorkchart package[L2]●

Owns domain intelligence and composes pipeline, algorithms, and protocols.

HAHuman gate[L3]●

Decides risky or irreversible transitions.

Instruments

API-S02Supported subpaths

Public contract communicates support level.

RUN-C01Typed pipeline[FR]

The canonical runtime for known multi-step paths.

TRU-G01Trust stack[FC]

Audit, metrics, eval, persistence, retry, step execution.

EVT-R01Trace evidence

Normalised events are the common runtime/test vocabulary.

Flows

Application/API/CLI consumer
supported import
Supported subpaths
Supported subpaths
factory
Workchart package
Workchart package
run
Typed pipeline
Typed pipeline
[Q]guard
Trust stack
Trust stack
receipt
Trace evidence
Trust stack
[risk]approval
Human gate
  • Use the least autonomous mechanism that satisfies the job.
  • Runtime and test-time speak the same trace vocabulary.
  • Root barrel compatibility remains transitional, not the supported API.

Agency bridge plan

bridge

The bridge does not rewrite agency. It makes rot visible, blocks hollow proof claims, standardises folders, migrates one real-value workflow at a time, then retires legacy surfaces.

Source: docs/architecture/libs/agency/03-bridge-plan.md

Agents

HAEngineer[L3]●

Executes the bridge through DB-native plan work.

DASales-dev workflow[L1]●

Best current typed-pipeline pattern; still needs trust-stack value proof.

OANext real-value workchart[L2]?

Marketing strategy or project plan before RFP relocation.

Instruments

MAP-M01Reality map

Keeps the work grounded in measured current state.

VAL-G02Value-proof gate[FC]

Blocks template outputs and hardcoded eval scores from counting as value.

STD-C01Folder boundary standard[FR]

Makes workchart, algorithm, routing, and bucket boundaries legible.

ACT-T01Capability activation

Moves proven capabilities behind product paths and receipts.

Flows

Reality map
[next]gate
Value-proof gate
Value-proof gate
shape
Folder boundary standard
Folder boundary standard
copy pattern
Sales-dev workflow
Sales-dev workflow
[proved]migrate
Next real-value workchart
Next real-value workchart
product path
Capability activation
Engineer
[L3]enforce
Value-proof gate
  • The next slice is value-proof gate plus folder-boundary cleanup.
  • Do not wire consumers to scaffold-only workflows.
  • Migrate RFP only after a lower-risk real-value path proves the pattern.

§05

Workchart exemplars

Typed agency pipelines rendered as ordered control schematics from agency-owned metadata.

Marketing strategy workchart

pipeline-exemplar

A typed pipeline where AI analysis and strategy generation are controlled by evidence-bearing eval standards before deterministic confidence calculation.

Source: libs/agency/src/lib/workcharts/marketing-strategy-workflow/marketing-strategy.pipeline.ts

MKT-V01Market Analysis
DAStrategy intelligence service[L1]●
aiMKT-MARKET-ANALYSIS-001

AI-backed analysis creates the context for every later decision.

analysis
MKT-C02Strategy Generation
DAStrategy intelligence service[L1]●
aiMKT-STRATEGY-GENERATION-001

Controller step turns analysis, objectives, channels, and constraints into a strategy.

strategy
MKT-T03Confidence Calculation
OAMarketing strategy pipeline[L2]●
compute

Deterministic transform produces the confidence receipt.

Feedback controls

strategy-generation
[eval]quality signal
market-analysis
  • AI work is visible as controlled instruments, not opaque service calls.
  • Eval standards are first-class evidence hooks.
  • The final confidence score is a deterministic receipt.

Sales-dev outreach workchart

pipeline-exemplar

A six-step pipeline that researches, scores, composes, evaluates, pauses for human review, and schedules outreach through explicit control points.

Source: libs/agency/src/lib/workcharts/sales-dev-outreach-workflow/outreach.pipeline.ts

SDR-M01Prospect Research
DAProspect intelligence ports[L1]●
aiSDR-PROSPECT-RESEARCH-001

Monitors public prospect context before scoring.

prospect
SDR-S02Lead Scoring
DAProspect intelligence ports[L1]●
aiSDR-LEAD-SCORING-001

Selector step classifies fit and tier.

tier
SDR-C03Outreach Composition
DAProspect intelligence ports[L1]●
aiSDR-OUTREACH-COMPOSITION-001

Controller step composes channel-specific outreach.

messages
SDR-V04SPCL Evaluation
DAProspect intelligence ports[L1]●
aiSDR-SPCL-QUALITY-001

Validator grades status, power, credibility, and likeness.

[medium risk]review
SDR-G05SPCL Quality Gate[FC]
HASME reviewer[L3]●
human-gatemedium risk

Fail-closed human gate can pause the pipeline before delivery.

[approved]release
SDR-T06Delivery Scheduling
OAOutreach pipeline[L2]●
compute

Transmits approved messages into a deterministic schedule receipt.

Feedback controls

quality-gate
[reject]revise
outreach-composition
  • The human gate is a visible control instrument, not hidden application logic.
  • Eval standards connect AI steps to proof semantics.
  • Scheduling only receives approved value flow.