Files
Jonathan Boniface 9e503885cc ALZ-F01 Discovery: package step, value scorecard, application grouping, cases
- New cp1-discovery-package-agent (ADR 0020): one package per reviewed run, every part
  hashed (sha256 over canonical JSON), with a manifest and a value scorecard.
- Applications grouped CMDB -> tag -> resource group with a judged model fallback (ADR 0021):
  CMDB 0.7.0 carries ownership, dependency mapping 0.6.0 groups, report 0.9.0 shows them.
- Critic 0.6.1 accepts the producers' enums; build_workflow.py refuses a consumer enum
  narrower than its producer's.
- Code hooks, sub-agents and tests for the inventory, CMDB, dependency, planner and critic
  agents; the report and Well-Architected review agents; ADRs 0012-0021.
- Workflow cases with value bars and run limits; run evaluation and usage collection scripts;
  the run context template and the field-by-field review for ADR 0016.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-05 22:13:25 +00:00

34 lines
2.2 KiB
Markdown

# CP1 - Well-Architected Review Agent
Reviews a live cloud estate against the six Well-Architected pillars in code: fixed checks on its configuration, the provider's own recommendations by pillar and impact, and guidance references, rendered as Markdown and HTML.
A step of the workflow "CP1 - Agentic Landing Zone - Discovery", after the critic and before the report. Like the
report agent, it answers from code (`app/waf.py`, called by the `pre_process` hook), so no model is involved and every
figure can be recomputed from the evidence. With `provider` `sample` it replies "not applicable" without calling a tool.
## What it reads
Only read-only tools on the existing servers, through the MCP gateway, with the project's stored connection:
`Azure Discovery` or `Google Cloud Discovery` (servers, workloads, clusters, data stores, firewall or NSG rules,
organisation policies, Azure resource locks, every page of the provider's recommendations, and one guidance reference
per pillar from the server's own `<provider>_search_well_architected`, which takes only fixed pillars and topics). The checks judge only the scope entries in `records`.
## What it writes
`waf_markdown` (the file), `waf_html` (the draft page) and `final_waf_html` (the page released after approval),
plus the structured `pillars`, `checks` and `recommendations`. A pillar with no evidence is **not assessed**, never a
pass. Calls that failed are listed in `call_issues` and in the review.
## Checks
| Pillar | Checks |
|---|---|
| Security | SEC-01 no admin or database port open to the internet; SEC-02 no public server address; SEC-03 Secure Boot on; SEC-04 storage blocks public access; SEC-05 guardrail policies at the scope root |
| Reliability | REL-01 compute not confined to one zone; REL-02 deletion protection (GCP) or resource locks (Azure); REL-03 database backups; REL-04 object versioning (GCP) |
| Cost | COST-01 no stopped server holding paid resources; COST-02 cost-allocation labels |
| Operations | OPS-01 owner or environment labels; OPS-02 storage access logging (GCP) |
| Sustainability | SUS-01 no idle or stopped resources |
| Performance | Provider recommendations only (no telemetry) |
Tests: `python3 -m pytest tests/` (no runtime needed).