Agent reference

This page is for AI agents (Cursor, Claude Code, Telemachus, MCP). Humans should start with Platform workflow or Core Concepts.

Fetch llms.txt and linked .md URLs — not HTML pages. Load one topic at a time; avoid llms-full.txt unless you need the full corpus.

How to read

Index: https://docs.rhesis.ai/llms.txt
Single page: e.g. https://docs.rhesis.ai/docs/metrics/metric-scope.md
Term lookup: https://docs.rhesis.ai/glossary/<term-id>.md (e.g. behavior, metric-scope, test-set)
Confusions only: skills/rhesis/references/definitions.md in the repo — when terms are mixed up, not for full definitions

User signal	Read first
Vague “help me test”	Platform workflow
Endpoint + explore	Architect exploration
Pasted PRD / FRs / requirements	Requirements workflow
Run / analyze	Execution and analysis
Terminology	Glossary first (`test-result`, `tag`, `mcp`); repo `definitions.md` for confusions only
`metric_scope`	Metric scope

Behavior vs metric — Behavior = what should happen. Metric = how you score it.
Category/topic vs behavior — Categories and topics organize tests; they do not replace behaviors or metrics.
PRD section title vs behavior — “Security” is not a behavior. Split numbered requirements into testable expectations.
Source vs pasted PRD — Chat paste is ephemeral; use create_source for large specs (Single-Turn only).

Table	Columns
Metrics	Metric, Behavior, AC source, `metric_scope`, Score type, Pass definition
Test sets	Test set, `test_type`, Behaviors, Metrics expected to run
Coverage matrix	Test set, test_type, Behavior, Linked metric, Scope OK?

Install: npx skills add rhesis-ai/rhesis -g

SKILL.md routes to these docs and to skills/rhesis/references/ in the repo. Connect the Rhesis MCP server for tool calls.

Loads the same skills/rhesis/references/ files per phase. See Architect workflow.

Fictional requirements plan shape: use-case-bracketfeld.md