feat: working TOM agent

This commit is contained in:
2026-10-21 19:47:12 +01:00
commit efe543e6e2
83 changed files with 11779 additions and 0 deletions

View File

@@ -0,0 +1,47 @@
---
name: operating-scoping
description: >-
What an operating spec's fields mean and how each is settled: the solution to be
operated, how it must be operated, who operates it, and in which currency its
cost is stated. Use when completing the operating scope, and whenever a
Deliverable must honour what the customer stated.
---
# The operating spec
An operating spec is the structured statement an Execution starts from. It holds what the customer wants from operations, the facts an operating model depends on, and the solution to be operated. `assets/catalogue.yaml` lists every field.
## What a field can say
| Written | Meaning | What you do |
| --- | --- | --- |
| a value | The customer states this | Use it as given. It binds the operating model. |
| `ask`, or the field left out | Not known yet | Ask when someone is present. |
| `decide` | You choose | Pick a value and record it as an assumption, with what it would change. |
| `not_applicable` | Does not apply | Ask and assume nothing. |
## Business outcomes
`intent.business_outcomes` is what the business wants from the solution, in the customer's words, each with an id (`BO-1`) and, where they give one, a target. Every measure in the operating model must show one of them, and each must be shown by at least one. Ask for them when someone is present: an operating model whose measures serve no stated outcome cannot show it is worth what it costs. Never invent one; one the customer did not state is not an outcome of theirs.
## The solution
- `solution.items` is what is to be operated. Every Item must be covered by a Capability. Never add an Item, remove one or rename one: it is the customer's description of their solution, or the digest of their architecture.
- `solution.conditions` is what must be kept true of an Item while it is run. Every Condition must be carried by a Check.
- `source_ref` on an Item or a Condition is its id where it came from, such as an Element or a Check of a solution architecture. Keep it.
- Something operations needs that the solution does not provide is a **finding**. Record it; do not design around it.
## What binds
A stated value of a field marked `binds` in the catalogue, and every requirement under `intent.requirements`, must be named as the `source` of an Operational requirement. What was assumed does not bind: it is shown apart and may change when the customer answers.
## Who operates
- `sourcing.model` says whether the customer, a service provider, or both operate the solution. It decides which Parties the operating model has.
- `sourcing.delivery_countries`, when stated, are the only countries a Role may sit in.
## Cost
- `commercial.currency` is the currency every cost is stated in. It is always settled before work begins.
- The customer's own staff are counted and never priced.
- Nobody states a rate in a spec. Rates are the service provider's, in the rate card Skill.

View File

@@ -0,0 +1,214 @@
# The operating spec's catalogue: every field an operating spec may hold.
#
# A field is a dotted path. `type` is one of: string, text, integer, number, boolean,
# date, list, enum, records. `tier` says when it must be settled: essential (always,
# before any work), important (asked when someone is present) or optional. A tier
# may differ by the kind of engagement. `question` is what to ask the customer for
# it. `default` guides the choice when the customer says "decide". `affects` says
# what a wrong assumption would change. `binds: true` marks a choice an operating
# model must honour: stated by the customer, it has to reach an Operational
# requirement.
#
# This catalogue names no Cloud provider and no currency. A provider's bundle adds
# the allowed values for the fields marked `provider_values`.
schema_version: 1
kinds: [workload, platform_foundation, migration]
fields:
metadata.customer_case:
type: string
tier: essential
aliases: [metadata.engagement_id]
question: What short reference identifies this engagement?
metadata.client_name:
type: string
metadata.author:
type: string
metadata.revision:
type: string
# Where the spec was digested from, when it was: written by the digest.
metadata.source:
type: string
engagement.kind:
type: enum
values: [workload, platform_foundation, migration]
tier: essential
question: >-
Is what is to be operated an application workload, a platform foundation
such as a landing zone, or systems being migrated?
intent.statement:
type: text
tier: essential
question: In the customer's own words, what do they want from operations?
intent.requirements:
type: records
keys: [id, statement]
optional_keys: [category]
question: Which requirements for operations has the customer stated outright?
intent.business_outcomes:
# What the business wants from the solution, in the customer's words. Every
# measure in the operating model must show one of these.
type: records
keys: [id, outcome]
optional_keys: [target, category]
# Essential: everything measured is tied to one, so they are settled before
# anything is designed. With nobody to ask, they are assumed and shown apart.
tier: essential
question: >-
What does the business want from this solution, and how would it know it was
getting it? Give each outcome a short id, such as BO-1.
default: >-
Take them from what the request says the solution is for. Two or three, each
something the business would notice, such as changes reaching customers
sooner or outages being short.
affects: what is measured, and what each measure is for
intent.exclusions:
type: list
tier: important
question: What has the customer said is out of scope for operations?
default: Nothing is excluded beyond what the statement implies.
affects: which Capabilities are designed
# --- the solution to be operated -----------------------------------------------
solution.cloud_provider:
type: enum
provider_values: true
tier: essential
question: Which Cloud provider does the solution run on?
solution.items:
type: records
keys: [id, name, kind]
# `elements` are the ids of the parts of a design an Item stands for;
# `evidence` is the passage a drafted Item was read from.
optional_keys: [product, environments, source_ref, elements, evidence]
tier: essential
question: What is to be operated? List each thing, with a short id and its kind.
solution.conditions:
type: records
# `on` is the id of the Item it is a condition on, or a list of them.
keys: [id, "on", condition]
optional_keys: [source_ref, evidence]
tier: important
question: What must be kept true of each thing while it is run?
default: None beyond what the service levels state.
affects: the Checks the operating model must carry
solution.build.duration_days:
type: integer
tier: important
question: How many working days is the solution's build planned to take?
default: Not known; the transition is planned without a build to fit.
affects: when Capabilities must be ready
solution.build.customer_owned:
type: list
question: Which parts of the build does the customer or their supplier own?
solution.inherited_risks:
type: records
keys: ["on", risk]
optional_keys: [source_ref]
question: What is already known to be unverified or at risk in the solution?
# --- how it is to be operated ---------------------------------------------------
operations.support_hours:
binds: true
type: enum
values: [business, extended, 24x7]
tier: essential
question: >-
During which hours must faults be responded to: business hours, extended
hours, or every hour of every day?
operations.availability_hours:
type: enum
values: [business, extended, 24x7]
tier: important
question: During which hours must the solution be available to its users?
default: The same as the support hours.
affects: whether support hours are wide enough
operations.change_process:
type: text
tier: important
question: How are changes approved and released today, and must that continue?
default: Changes go through a pipeline with peer review and no separate board.
affects: change enablement and release
operations.monitoring:
type: text
question: Which monitoring and alerting is already in place or required?
sourcing.model:
binds: true
type: enum
values: [customer_operated, provider_operated, shared]
tier: essential
question: >-
Who operates the solution: the customer, a service provider, or the two
together?
sourcing.delivery_countries:
binds: true
type: list
tier: important
question: From which countries may the solution be operated? Use country codes.
default: No restriction beyond where the data may be handled.
affects: where Roles sit, and the cost
organisation.existing_teams:
type: records
keys: [name, party]
optional_keys: [skills, hours, country]
tier: {migration: essential, workload: important, platform_foundation: important}
question: Which teams exist today that will have a part in operating this?
default: None exist; every Role is new.
affects: which Roles are new and which are filled
organisation.headcount_limit:
type: integer
question: Is there a limit on the number of people operating this?
service_management.tooling:
type: text
tier: important
question: Which service management and on-call tooling must be used?
default: The Cloud provider's own operations tooling, with no separate system.
affects: tooling and the processes written for it
service_levels.priority_1_response_minutes:
binds: true
type: integer
tier: important
question: Within how many minutes must the most serious incident be acknowledged?
default: 30
affects: on-call arrangements and staffing
service_levels.recovery_time_hours:
binds: true
type: number
tier: important
question: Within how many hours must service be restored after a major failure?
default: 4
affects: continuity arrangements
compliance.frameworks:
type: list
tier: important
question: Which regulations or standards must operations comply with?
default: None beyond the organisation's own policies.
affects: which Capabilities apply and what they must evidence
commercial.currency:
type: enum
provider_values: true
tier: essential
question: In which currency should the cost of operating be stated?
commercial.monthly_run_budget:
type: number
question: Is there a monthly budget for operating this, in that currency?
transition.go_live_date:
type: date
question: When must the solution be in operation?
# --- how to run the agent -------------------------------------------------------
run.pre_approved:
type: boolean
question: Should the agent run without stopping for approvals?
run.overlay_skills:
type: list
question: Which overlay Skills should this Execution load?