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 +1,50 @@
---
name: operating-capabilities
description: >-
The capability catalogue: what an operating organisation may need to be able to
do, in our own words, with the industry frameworks each entry draws on. Use when
choosing the Capabilities of an operating model and writing the Checks on them.
---
# The capability catalogue
`assets/catalogue.yaml` lists the Capabilities an operating model is built from. `assets/sources.yaml` is the register of the frameworks they cite.
## How to use it
1. Your task's context lists the entries code found to apply to this operating spec, under `capabilities_proposed`. Start from those.
2. Take every proposed entry, or say in your Deliverable why one does not apply. Add any other entry the solution needs.
3. A Capability names its entry by `catalogue`. One the catalogue does not have is `custom`, with a `reason`: use it only when no entry fits, never to rename one.
4. Give each Capability the Items it covers and the Operational requirements it serves. Every Item must be covered by at least one Capability, and every Operational requirement served by at least one.
5. An entry's `typical_checks` are where to start when writing Checks. Rewrite each for this solution: name the Item, and put the customer's own numbers in.
## What the sources are for
A source says where an idea comes from. Cite it in a Deliverable as the catalogue does, by framework and by the name of the part. Never quote a framework's text and never say an operating model is certified against, or compliant with, a framework: the catalogue draws on them and promises nothing about them.
The register's `status` says whether a person has checked the references. While it is `unverified`, say so wherever a source is cited to a customer.
## People and teams
Two frameworks shape who does the work, not only what it is:
- **Team Topologies** (`team-topologies`). An operating model for a platform may group its Roles into teams, each of one of four kinds: `stream_aligned`, `platform`, `enabling` or `complicated_subsystem`. Two teams work together in one of three modes: `collaboration`, `x_as_a_service` or `facilitating`. A collaboration is a stage for discovering something together, not a standing arrangement, so it says when it ends.
- **Platform as a product** (`platform-as-a-product`). A platform is run for the teams that build on it, as a product they choose to use: someone owns its roadmap and orders the backlog by what slows those teams down, adoption is earned by a supported default path, and their satisfaction is measured.
For a platform foundation, the catalogue proposes `platform-product-management`, `self-service-and-golden-paths` and `enablement` from these.
## Measuring at three layers
`delivery-and-developer-experience-measurement` draws on three frameworks at once, because each measures a different layer and none stands in for another:
| Layer | Framework | What it tells you |
| --- | --- | --- |
| Outcome | `dora-metrics` | How fast and how safely changes reach production: lagging. |
| Guardrail | `space-framework` | Whether productivity is sustainable across several dimensions, including wellbeing. |
| Leading signal | `devex-framework` | The friction, feedback loops and cognitive load teams meet: what to improve next. |
Never write a Check that measures only one layer and calls it productivity.
Every one of these is a means. In the operating model each is a **Measure** tied to a business outcome the customer stated: deployment frequency to "new products reach customers sooner", developer satisfaction to "we keep the engineers we hire", time to restore to "payments are never down for long". A measure that serves no stated outcome is not measured.
Two things are called DORA. `eu-dora` is the European regulation on operational resilience. `dora-metrics` is the research programme's four delivery measures. Name each in full.

View File

@@ -0,0 +1,281 @@
# The capability catalogue: what an operating organisation may need to be able to do.
#
# Every entry is in our own words and cites the frameworks it draws on, each of
# which must be in sources.yaml. `domain` groups entries. `applies_when` says when
# code proposes the entry for an operating spec:
# always: true every operating model
# item_kinds: [database, storage] when the spec lists an Item of one of these kinds
# compliance: [DORA] when the spec names one of these frameworks
# sourcing: [shared] for these sourcing models
# engagement_kinds: [platform_foundation] for these kinds of engagement
# `follows_support_hours: true` marks a capability that must be staffed for at
# least the hours the spec says faults are responded to.
schema_version: 1
capabilities:
- id: incident-management
name: Incident management
domain: service_operations
purpose: Restore a service that is degraded or down, and say what happened.
applies_when: {always: true}
follows_support_hours: true
typical_checks:
- The most serious incident is acknowledged within the agreed time, at any hour faults are responded to.
- Every incident of the most serious kind has a written review.
sources:
- {framework: itil-4, reference: Incident management practice}
- {framework: google-sre, reference: Managing Incidents}
- {framework: iso-20000-1, reference: Incident management}
- {framework: eu-dora, reference: "Article 17, ICT-related incident management process"}
- id: monitoring-and-alerting
name: Monitoring and alerting
domain: service_operations
purpose: Know the state of what is run, and tell the right people when it needs them.
applies_when: {always: true}
follows_support_hours: true
typical_checks:
- Every Item has a signal that shows whether it is serving, and an alert when it is not.
- Every alert names who receives it and what they do.
sources:
- {framework: itil-4, reference: Monitoring and event management practice}
- {framework: google-sre, reference: Monitoring Distributed Systems}
- {framework: nist-csf-2, reference: "Detect: Continuous Monitoring"}
- {framework: eu-dora, reference: "Article 10, Detection"}
- id: problem-management
name: Problem management
domain: service_operations
purpose: Find why incidents happen and remove the cause.
applies_when: {always: true}
typical_checks:
- A recurring incident has a named cause and an owner for removing it.
sources:
- {framework: itil-4, reference: Problem management practice}
- {framework: google-sre, reference: "Postmortem Culture: Learning from Failure"}
- id: service-request-fulfilment
name: Service request fulfilment
domain: service_operations
purpose: Handle what users and teams ask for that is not a fault, such as access or a new environment.
applies_when: {always: true}
typical_checks:
- Each kind of request has a stated way to ask and a stated time to fulfil.
sources:
- {framework: itil-4, reference: Service request management practice}
- {framework: iso-20000-1, reference: Service request management}
- id: change-enablement
name: Change enablement
domain: change
purpose: Make changes to what is run with the risk of each understood and accepted.
applies_when: {always: true}
typical_checks:
- Every change to a production Item is reviewed by someone other than its author before it is applied.
- Every change can be traced to who approved it.
sources:
- {framework: itil-4, reference: Change enablement practice}
- {framework: iso-20000-1, reference: Change management}
- {framework: dora-metrics, reference: Change failure rate}
- id: release-and-deployment
name: Release and deployment
domain: change
purpose: Put approved changes into service repeatably, and take them out again when they fail.
applies_when: {always: true}
typical_checks:
- A failed deployment can be rolled back within the agreed recovery time.
sources:
- {framework: itil-4, reference: Release management and Deployment management practices}
- {framework: google-sre, reference: Release Engineering}
- {framework: dora-metrics, reference: Deployment frequency and lead time for changes}
- id: configuration-management
name: Configuration management
domain: change
purpose: Know what is run, how it is configured, and that it matches what was approved.
applies_when: {always: true}
typical_checks:
- The configuration of every Item is held in version control and compared with what is running.
sources:
- {framework: itil-4, reference: Service configuration management practice}
- {framework: nist-csf-2, reference: "Identify: Asset Management"}
- id: backup-and-recovery
name: Backup and recovery
domain: resilience
purpose: Keep copies of what cannot be rebuilt, and prove they can be restored.
applies_when: {item_kinds: [database, storage, data]}
typical_checks:
- A restore of every data store is tested on a schedule, and the result recorded.
sources:
- {framework: itil-4, reference: Service continuity management practice}
- {framework: nist-csf-2, reference: "Recover: Incident Recovery Plan Execution"}
- {framework: eu-dora, reference: "Article 12, Backup policies and procedures, restoration and recovery procedures and methods"}
- id: service-continuity
name: Service continuity
domain: resilience
purpose: Restore service after the loss of a site, a region or a supplier, within the agreed time.
applies_when: {always: true}
typical_checks:
- Recovery from the loss of the primary location is rehearsed, and takes no longer than the agreed recovery time.
sources:
- {framework: itil-4, reference: Service continuity management practice}
- {framework: iso-20000-1, reference: Service continuity management}
- {framework: eu-dora, reference: "Article 11, Response and recovery"}
- id: capacity-and-performance
name: Capacity and performance management
domain: resilience
purpose: Keep enough capacity for demand, and no more than is paid for knowingly.
applies_when: {always: true}
typical_checks:
- Each Item's use is reviewed against its limits on a schedule.
sources:
- {framework: itil-4, reference: Capacity and performance management practice}
- {framework: iso-20000-1, reference: Capacity management}
- id: access-management
name: Access management
domain: security_operations
purpose: Give people and systems the access they need, and take it away when they no longer do.
applies_when: {always: true}
typical_checks:
- Access to production is granted for a stated reason and reviewed on a schedule.
- The team that responds to incidents has a tested way in at any hour it is on call.
sources:
- {framework: nist-csf-2, reference: "Protect: Identity Management, Authentication, and Access Control"}
- {framework: iso-27001, reference: Access control}
- {framework: eu-dora, reference: "Article 9, Protection and prevention"}
- id: vulnerability-and-patch-management
name: Vulnerability and patch management
domain: security_operations
purpose: Find weaknesses in what is run and remove them within an agreed time.
applies_when: {always: true}
typical_checks:
- A weakness of the most serious kind is fixed or accepted in writing within the agreed time.
sources:
- {framework: nist-csf-2, reference: "Protect: Platform Security"}
- {framework: iso-27001, reference: Management of technical vulnerabilities}
- {framework: eu-dora, reference: "Article 9, Protection and prevention"}
- id: security-monitoring-and-response
name: Security monitoring and response
domain: security_operations
purpose: Detect misuse of what is run and respond to it.
applies_when: {always: true}
follows_support_hours: true
typical_checks:
- Security events from every Item reach the people who watch for them.
sources:
- {framework: nist-csf-2, reference: "Detect: Adverse Event Analysis; Respond: Incident Management"}
- {framework: iso-27001, reference: Information security incident management}
- {framework: eu-dora, reference: "Article 10, Detection"}
- id: cost-management
name: Cost management
domain: financial_operations
purpose: Know what running costs, who it is spent for, and when it departs from plan.
applies_when: {always: true}
typical_checks:
- Spend is reported against budget monthly, by Item.
- Spend that departs from forecast raises an alert to a named Role.
sources:
- {framework: finops-framework, reference: "Allocation; Budgeting; Forecasting; Anomaly Management"}
- id: service-level-management
name: Service level management
domain: governance
purpose: Agree what the service must achieve, measure it, and report it.
applies_when: {always: true}
typical_checks:
- Each service level the customer stated is measured and reported on a schedule.
sources:
- {framework: itil-4, reference: Service level management practice}
- {framework: google-sre, reference: Service Level Objectives}
- {framework: iso-20000-1, reference: Service level management}
- id: platform-product-management
name: Platform product management
domain: platform_product
purpose: >-
Run the platform as a product for the teams that build on it: learn what slows
them down, keep a roadmap and a backlog ordered by that friction, and measure
whether they choose to use it.
applies_when: {engagement_kinds: [platform_foundation]}
typical_checks:
- A named Role owns the platform's roadmap and orders its backlog by what slows the teams that use it.
- Adoption and the satisfaction of the teams that use the platform are measured on a schedule and reported.
sources:
- {framework: platform-as-a-product, reference: Platform as a product; Measuring platform success}
- {framework: team-topologies, reference: Platform team; Thinnest viable platform}
- id: self-service-and-golden-paths
name: Self-service and golden paths
domain: platform_product
purpose: >-
Offer what the teams building on the platform need as self-service, with a
supported default way to do the common things, so that using it is the
easier choice rather than a rule.
applies_when: {engagement_kinds: [platform_foundation]}
typical_checks:
- The common requests of the teams that use the platform are met through self-service, with no ticket.
- Every self-service path is documented and has a named owner.
sources:
- {framework: team-topologies, reference: X-as-a-Service interaction mode}
- {framework: platform-as-a-product, reference: Platform capabilities; Self-service}
- id: enablement
name: Enablement
domain: platform_product
purpose: >-
Help the teams that build on the platform adopt it, by pairing, pilots,
documentation and training, for a stated time, rather than by mandate.
applies_when: {engagement_kinds: [platform_foundation]}
typical_checks:
- Each engagement to help a team adopt the platform has a stated goal and end date.
sources:
- {framework: team-topologies, reference: Enabling team; Facilitating interaction mode}
- {framework: platform-as-a-product, reference: Platform as a product}
- id: delivery-and-developer-experience-measurement
name: Delivery and developer experience measurement
domain: governance
purpose: >-
Measure how well work flows through the platform at three layers that answer
different questions: delivery speed and stability, the outcome; productivity
and wellbeing across the team, the guardrail; and the friction teams meet,
the leading signal of what to improve next. No one layer stands in for another.
applies_when: {engagement_kinds: [platform_foundation, workload]}
typical_checks:
- Deployment frequency, lead time for changes, change failure rate and time to restore service are measured and reported monthly.
- Productivity is reviewed across more than one dimension, including satisfaction and wellbeing, and never by an activity count alone.
- The friction teams meet, their feedback loops and their cognitive load, are surveyed on a schedule and set the order of the platform's backlog.
sources:
- {framework: dora-metrics, reference: Deployment frequency; Lead time for changes; Change failure rate; Time to restore service}
- {framework: space-framework, reference: "Satisfaction and well-being; Performance; Activity; Communication and collaboration; Efficiency and flow"}
- {framework: devex-framework, reference: Feedback loops; Cognitive load; Flow state}
- id: supplier-management
name: Supplier management
domain: governance
purpose: Hold the suppliers the service depends on to what they agreed.
applies_when: {sourcing: [provider_operated, shared]}
typical_checks:
- Each supplier the service depends on has a named owner and a way to escalate.
sources:
- {framework: itil-4, reference: Supplier management practice}
- {framework: iso-20000-1, reference: Supplier management}
- {framework: eu-dora, reference: "Article 28, General principles on ICT third-party risk"}
- id: regulatory-evidence
name: Regulatory evidence and reporting
domain: governance
purpose: Keep the evidence a regulation asks for and report what it requires, when it requires it.
applies_when: {compliance: [DORA, PCI_DSS_v4, ISO27001, ISO_27001, SOC_2]}
typical_checks:
- Evidence each named framework asks for is kept for as long as it asks.
sources:
- {framework: eu-dora, reference: "Article 19, Reporting of major ICT-related incidents"}
- {framework: iso-27001, reference: Compliance with legal and contractual requirements}

View File

@@ -0,0 +1,82 @@
# The register of frameworks the capability catalogue cites.
#
# A capability's entry is written in our own words. A framework is named as a
# source, by the title of the part drawn on, and never quoted. `status` says
# whether a person has checked the entry's references against the framework:
# every reference here was drafted from general knowledge of the framework and is
# `unverified` until someone has.
schema_version: 1
status: unverified
frameworks:
itil-4:
name: ITIL 4
publisher: PeopleCert
version: "4"
licence: Proprietary; cited by practice name, not reproduced.
google-sre:
name: "Site Reliability Engineering: How Google Runs Production Systems"
publisher: Google / O'Reilly
version: "2016"
licence: Published under a no-derivatives licence; cited by chapter title, not reproduced.
finops-framework:
name: FinOps Framework
publisher: FinOps Foundation
version: "2024"
licence: Openly licensed with attribution; cited by capability name.
nist-csf-2:
name: NIST Cybersecurity Framework
publisher: National Institute of Standards and Technology
version: "2.0"
licence: Public domain; cited by function and category.
iso-20000-1:
name: ISO/IEC 20000-1
publisher: ISO/IEC
version: "2018"
licence: Proprietary; cited by requirement title, not reproduced.
iso-27001:
name: ISO/IEC 27001
publisher: ISO/IEC
version: "2022"
licence: Proprietary; cited by control theme, not reproduced.
eu-dora:
# The EU regulation. Not the DevOps Research and Assessment programme's metrics.
name: Regulation (EU) 2022/2554, the Digital Operational Resilience Act
publisher: European Union
version: "2022"
licence: Public law; cited by article.
dora-metrics:
# The research programme's delivery metrics. Not the EU regulation above.
name: DevOps Research and Assessment (DORA) software delivery metrics
publisher: Google Cloud
version: "2023"
licence: Publicly published research; cited by metric name.
# Socio-technical and organisational frameworks: how people, teams and
# responsibilities are structured around what is operated.
team-topologies:
name: "Team Topologies: Organizing Business and Technology Teams for Fast Flow"
publisher: Matthew Skelton and Manuel Pais, IT Revolution
version: "2019"
licence: Proprietary; cited by the names of its team types and interaction modes, not reproduced.
platform-as-a-product:
# The practice of running an internal platform as a product for the teams that
# use it. Cited from the CNCF's write-up of it, the most widely referenced one.
name: CNCF Platforms White Paper (platform as a product)
publisher: Cloud Native Computing Foundation, TAG App Delivery
version: "2023"
licence: Openly published; its licence is to be confirmed; cited by section, not reproduced.
# Measuring how well the work flows, at three layers. None replaces another.
space-framework:
name: "The SPACE of Developer Productivity"
publisher: Forsgren, Storey, Maddila, Zimmermann, Houck and Butler, ACM Queue
version: "2021"
licence: Published article; cited by its five dimensions, not reproduced.
devex-framework:
name: "DevEx: What Actually Drives Productivity"
publisher: Noda, Storey, Forsgren and Greiler, ACM Queue
version: "2023"
licence: Published article; cited by its three dimensions, not reproduced.
sfia:
name: Skills Framework for the Information Age
publisher: SFIA Foundation
version: "9"
licence: Proprietary; cited by skill name, not reproduced.