sgit newsroom v0.1.29 · snapshot 2026-09-24

Reading room · teams.sgit.ai · llms-full

On this page

Reading room / teams.sgit.ai / llms-full.txt · section 24 of 33

01 — The ROLE.md Format: measured, and one recommendation

Version v0.33.64 · 7 September 2026 Method All 39 ROLE.md files on disk parsed for section headers and Identity fields, 7 September 2026. Every count below is generated, not estimated.


1. The schema

Five fields make up a role's identity, and they appear (in one markup or the other) across 31 of the 39 files:

FieldWhat it doesWhy it matters
NameThe role's handleIt is how other roles address it in comms
Locationteam/roles/<name>/The role's home for reviews and outputs
Core MissionOne sentence of purposeThe routing key — the Conductor reads this to assign work
Central ClaimThe role's testable assertionSee §3. This is the field that does the real work
Not Responsible ForThe explicit exclusion listSee §2. This is the field nobody else writes

2. Not Responsible For — the field that makes it a team

31 of 39 role files carry it. It is the most distinctive thing in the corpus and the reason a multi-role setup works at all.

Conductor: "writing code, running tests, deploying infrastructure, making architecture decisions, or performing security reviews." Librarian: "writing application code, making architecture decisions, running tests, deploying infrastructure, creating original specifications, or making product decisions."

Why it matters more than the responsibility list: an LLM given a task will attempt it. Capability is not the constraint — willingness is, and the exclusion list is the only thing that converts a capable generalist into a specialist that hands off. Without it, every role silently becomes the same role, and a "team" of nine agents is one agent invoked nine times.

The site should state this as a rule: a role without an exclusion list is not a role.

3. Central Claim — write it as a failure condition

This is the pack's most useful finding, and it is a regression the estate has not noticed.

The older bullet-list format states claims as falsifiable failures:

The newer table format states them descriptively:

The descriptive claims are true and useful. But you cannot check them. The failure-condition claims name a condition, a threshold and sometimes a time bound — thirty seconds, sixty seconds — which means an auditor (human or agent) can look for a counter-example and find one. That is the same falsifiability discipline that runs through risks.sgit.ai, wardley-maps.sgit.ai ("maps are claims") and threat-modeling.sgit.ai (the validated threat model), applied to organisational design.

Recommendation for the site: publish the failure-condition form as the canonical one, show the drift honestly, and offer a rewrite of the seven descriptive claims into testable form as an open build item.

4. Format drift, quantified

Across the 17 Explorer role directories:

StateCountRoles
Identity as table (| **Field** |)7architect, conductor, designer, dev, devops, dpo, qa
Identity as bullet list (- **Field:**)6appsec, cartographer, grc, historian, journalist, librarian
No ROLE.md at all4advocate, alchemist, ambassador, sherpa

Two markup dialects of one schema, plus four directories that exist without a definition. The schema is stable; the presentation is not. A parser reading these files must handle both — which is itself an argument for the site publishing a canonical machine-readable form (teams__roster.json in this pack is the first cut) rather than only prose.

5. Section anatomy, by frequency

Measured across all 39 files:

SectionFilesNote
## Identity37The schema above
## Tools and Access37Where skills are named — link out to skills.sgit.ai
## For AI Agents37The role speaking to its own occupant — see below
## Quality Gates32What must be true before the role signs off
## Core Workflows25The role's repeatable procedures
## Primary Responsibilities19
## Integration with Other Roles19The routing table, per role
## Measuring Effectiveness18
## Escalation18When to hand up rather than out
## What You DO (Villager Mode)17Topology-specific mandate
## What You Do NOT Do15The prose form of the exclusion list
## Incident Response12

## For AI Agents in 37 of 39 files is the quiet innovation. These documents are written for a non-human occupant and say so in a dedicated section. That is the difference between an org chart and an agent brief, and it is the section a reader building their own team should copy first.


This document is released under the Creative Commons Attribution 4.0 International licence (CC BY 4.0).


source: /briefs/02__the-roster.md