Files
gcp_solution_architecture_a…/CONTEXT.md
Jonathan Boniface c9f59be274 docs: accept ADR-0004, lessons and overlays arrive as Skills
Accepted by the owner on 2026-10-08 with its six decisions as written:
a person approves every lesson; a lesson must be shown to help by
evaluation; lessons are per Cloud provider; overlay Skills come only from
configured directories; each Execution records the Skills it loaded; and
lesson candidates are client data kept out of the repository.

- CONTEXT.md gains Lesson candidate, Lesson and Overlay Skill.
- Iteration 4 step 9 is complete; steps 10 and 12 name the decisions they
  must meet and are unblocked. Neither is built.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-08 15:05:03 +01:00

6.9 KiB

Cloud Solution Architecture

Turns a request for a cloud workload into a reviewed solution architecture: requirements, a target design with diagram and infrastructure-as-code, validation results, and a packaged guide. Google Cloud is the first Cloud provider.

Language

Running the workflow

Workflow request: The free-text description of the workload a caller wants architected; the opening input to an Execution. Avoid: prompt, query, request summary

Execution: One solution-architecture engagement for a single Workflow request. It lasts from the first request to an approved Solution guide, and may span many user turns. An Execution starts clean, unless it is a Change request. Avoid: run, job, workflow execution

Session: A caller-supplied grouping of Executions. A Session never feeds earlier outputs into a later Execution. Avoid: conversation, context

Customer case: A caller-supplied grouping of the Executions run for one customer workload, identified by a short slug. Every Execution belongs to one; an Execution started without one belongs to unassigned. A Customer case ties each request to the Deliverables generated for it. It feeds one Execution's outputs into another only when the caller starts a Change request and names the Baseline. Avoid: engagement, project, account, tenant

Intent: What the customer wants from a Customer case, in their own words and at a high level. Each Workflow request made in the case is a statement of it, and the statements are kept in order as its history. The Intent is the customer's; requirements are what the agents derive from it. Avoid: brief, ask, requirements (which are the derived Deliverable)

Change request: An Execution started to say how the customer's Intent has changed. It begins from a Baseline and delivers the full updated Deliverables with an account of what changed and why. Avoid: amendment, update run, iteration, re-run

Baseline: The completed Execution, in the same Customer case, that a Change request starts from. A Baseline is never modified. Avoid: parent, previous run, source Execution

Phase: A named grouping of Deliverables: requirements discovery, solution architecture (the design, ending with the Target architecture), solution implementation, solution validation, or packaging. A Phase is complete when all its Deliverables are submitted and the user has granted the Phase's Approval. Avoid: step, node, stage, pipeline

Approval: A user's explicit acceptance of a Phase's Deliverables, required before work that depends on them may start. An Execution is interactive by default. A pre-approved Execution treats its Approvals as already granted. Avoid: sign-off, confirmation, gate (in prose)

Revision: Reopening a Deliverable because validation found a problem or the user asked for a change. It returns that Deliverable and everything that depends on it to draft, and withdraws any Approval that covered them. Revisions per Deliverable are limited, and past the limit the user decides. Avoid: review round, retry, loop iteration

Agents

Orchestrator: The top-level agent of an Execution, and the only agent that talks to the user. It reasons about which Deliverables are ready to produce, delegates them to Specialists, and asks the user for Approvals. Avoid: supervisor, manager, root agent, router

Specialist: An agent responsible for producing one kind of Deliverable, such as the product selector or the target architect. A Specialist may delegate independent parts of its work to further Specialists. Avoid: sub-agent, worker, phase agent, node

Portability

Skill: A self-contained package of domain guidance: instructions plus optional references, assets and scripts, in the Agent Skills format. It is the portable unit of the agent's behaviour and is loaded on demand. Avoid: prompt template, instructions, playbook

Cloud provider: The hyperscaler a solution architecture targets, such as Google Cloud. Each Cloud provider brings its own Skills, plus the MCP servers it hosts for discovering a user's current environment and grounding recommendations in its documentation. Avoid: hyperscaler, cloud, platform, target

Lesson candidate: Something an Execution got wrong or was corrected on, recorded as it happened: a Revision and its reason, a tool refusal, a user's request for changes. It is raw material and changes nothing by itself. Avoid: learning, memory, feedback record

Lesson: A rule distilled from Lesson candidates across Executions, approved by a person and shown by evaluation to help, then published in a Skill. It carries no client detail. Avoid: memory, fine-tuning, learned behaviour

Overlay Skill: A Skill that adds an organisation's standards or one engagement's constraints to what the Cloud provider's Skills say. It is named when an Execution starts and comes only from a place the deployment has configured. Avoid: custom prompt, override, plugin

Outputs

Deliverable: A document an Execution produces: requirements, technical decomposition, product selection, design recommendations, the Target architecture, architecture description, deployment guidance, validation results, or the Solution guide. Each Deliverable depends on others being approved first. Avoid: artifact, output file

Dependency: The relationship between two Deliverables in which one can only be produced once the other has been accepted, for example an architecture diagram depending on the product selection. Within a Phase, a submitted Deliverable counts as accepted. Across Phases, only the earlier Phase's Approval does. Avoid: prerequisite, edge, upstream

Target architecture: The solution to be built, as one explicit graph of Elements, the Relationships between them, and the Checks each must meet. It is established before any infrastructure code is written, and every later Deliverable is derived from it and checked against it. Avoid: build plan, blueprint, design doc, to-be state

Element: One component of the Target architecture and the cloud resource that realises it, with the product selected for it. Avoid: node, resource (alone), box

Relationship: A connection between two Elements, such as calls, publishes to, reads from, or is granted access to. Avoid: edge, link, dependency (which is between Deliverables)

Check: A numbered, verifiable condition that an Element or Relationship must meet, traced to the requirement or design recommendation it comes from. Validation gives each Check a verdict. Avoid: finding, rule, control, test

Requirements chain: The unbroken trace from each requirement, through the components that serve it and the Elements and Checks that realise it, to the code that implements them and the validation verdict on each Check. Avoid: traceability matrix, lineage

Solution guide: The packaged Deliverable that assembles every other Deliverable into the Cloud provider's solution architecture template. Avoid: final report, package