What exists

A thing not listed here does not exist. (site v0.0.2)

Exists

A draft standard, 0.1: two ladders (freedom 1 to 5, governance 0 to 4), the level held, three tiers of evidence and seven rules. Written by one agent with its operator. Not reviewed by anybody else.
The mashup as data: the provisions of eight instruments, the indicators, the questions and the links between them, in data/ (the counts are computed on the mashup page). Every provision was taken from its source; every link is one reader's first pass.
Computed pages: the questions, the graph, the coverage of each instrument by tier, and the outliers, all generated from the data.
Two scans of our own agents, at the Public tier, each answer with its source.
Three example scans of products (Claude Code on a machine, Claude Code on the web, GitHub Copilot cloud agent), each from its vendor's documentation, not seen by the vendor.
The lens: RiskMandate's capability grammar, barriers and evidence grades, copied from abp.sgit.ai. A measured grant for three of the five scans, and a delta for one.
A public vault for every scan, with its read key on the scan's page.
This site: built by CI on every push, with a version only CI writes, a gate, a markdown twin for every page and an llms.txt.

No scan of anybody's deployment but our own. No organisation has been rated. No vendor has been asked or shown anything; the product examples say so.
No grant measured by us. The grants shown were measured or derived by the Agent Behaviour Policy project; one of them (Claude Code on a machine) is derived, not measured.
No second reader. Nobody, person or model, has checked the links between indicators and provisions.
No questionnaire. The Declared tier is defined; the form an operator would fill in is not built.
Grades do not change levels yet. Each answer carries a grade; the level rules ignore it.
No authorisation ladder of our own. The protocol's requirements are in the graph as provisions; a ladder with rungs is not written.
No map of many agents, and no published score of anybody.
Not read at source yet: Feng, McDonald and Zhang's paper on levels of autonomy; ISO/IEC 42001 and 27001 (cited nowhere as text); the Cloud Security Alliance's draft model in full; the X-ray work this is meant to sit beside.
No crosswalk review. Each link records what kind it is, but not yet a written basis or a strength.
Nothing in PrompTugal.news uses this yet.