docs(triage): the role axis is eight seats plus a non-seat, not six #1074

Merged

coilysiren merged 1 commit from aos/claude/ym96-role-axis into main

2026-08-15 22:02:05 +00:00

coilyco-ops commented

2026-08-15 21:59:45 +00:00

Member

Follows the label work in #1067 and #1069. The role/* axis grew twice today and the skill still described the original five.

What changed upstream

role/ops, then role/director and role/exec. And role/human was retuned from "requires a human to physically do something" to "requires a person, and specifically not an agent seat" - which turns it from a badly-fitting seat into an honest catch-all.

What this records

The table now lists all nine values.
How the axis arrived, in two rounds, because the sequence is the argument for it. The first round shipped ai, creator, design, engineer, qa, and role/human inherited everything else. That put a director's decisions and an operator's actions in one pile, so ops, director and exec followed.
A decision with a landing carries two seats - the one that rules and the one that applies the result. That pattern showed up immediately: raising a completion ceiling is director plus ops.

The evidence, from the backfill rather than from theory

Backfilling coilyco-gaming put 29 issues on role/human before the later values existed. Eighteen turned out to be operator actions and eleven were decisions. Once each had a real seat, role/human emptied entirely - 148 open issues, none of them needing it.

That is the sharpest available argument for the axis: a value holding everything unassigned tells you nothing about whose queue an issue is in, which is the only question role/* exists to answer.

On the size constraint

#1070 noted COMPOSED.md sat at 3998 of 4000 body characters and that the next addition there should displace rather than extend. So the entrypoint's role paragraph is rewritten shorter (3991 now) and the substance lives in label-taxonomy.md. Nothing else in the entrypoint moved.

pre-commit clean on the changed files, contract tests pass.

🤖 Generated with Claude Code

Follows the label work in #1067 and #1069. The `role/*` axis grew twice today and the skill still described the original five. ## What changed upstream `role/ops`, then `role/director` and `role/exec`. And `role/human` was retuned from *"requires a human to physically do something"* to *"requires a person, and specifically not an agent seat"* - which turns it from a badly-fitting seat into an honest catch-all. ## What this records * The table now lists all nine values. * **How the axis arrived, in two rounds**, because the sequence is the argument for it. The first round shipped `ai`, `creator`, `design`, `engineer`, `qa`, and `role/human` inherited everything else. That put a director's decisions and an operator's actions in one pile, so `ops`, `director` and `exec` followed. * **A decision with a landing carries two seats** - the one that rules and the one that applies the result. That pattern showed up immediately: raising a completion ceiling is `director` plus `ops`. ## The evidence, from the backfill rather than from theory Backfilling coilyco-gaming put **29 issues on `role/human`** before the later values existed. Eighteen turned out to be operator actions and eleven were decisions. Once each had a real seat, **`role/human` emptied entirely** - 148 open issues, none of them needing it. That is the sharpest available argument for the axis: a value holding everything unassigned tells you nothing about whose queue an issue is in, which is the only question `role/*` exists to answer. ## On the size constraint #1070 noted COMPOSED.md sat at 3998 of 4000 body characters and that the next addition there should displace rather than extend. So the entrypoint's role paragraph is **rewritten shorter** (3991 now) and the substance lives in `label-taxonomy.md`. Nothing else in the entrypoint moved. `pre-commit` clean on the changed files, contract tests pass. 🤖 Generated with [Claude Code](https://claude.com/claude-code)

coilyco-ops added 1 commit

2026-08-15 21:59:45 +00:00

docs(triage): the role axis is eight seats plus a non-seat, not six

ci / ward-doctor (pull_request) Successful in 13s

Details

ci / aos-cli-tests (pull_request) Successful in 18s

Details

ci / gate (pull_request) Successful in 48s

Details

ede406719d

Kai added role/ops, then role/director and role/exec, and retuned role/human
from "requires a human to physically do something" to "requires a person, and
specifically not an agent seat". The skill still described the first five.

label-taxonomy carries the change: the table, the two rounds the axis arrived
in, and why role/human stopped being a seat. The entrypoint's role paragraph is
rewritten rather than extended, because #1070 left COMPOSED.md at 3998 of 4000
body characters and said the next addition should displace.

The evidence for why the later values were needed comes from backfilling one
org: twenty-nine issues landed on role/human before they existed, eighteen were
operator actions and eleven were decisions, and role/human emptied once each
had a real seat. A value that holds everything unassigned answers nothing about
whose queue an issue is in, which is the axis's only job.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: Kai Siren <coilysiren@gmail.com>
Co-authored-by: Claude <noreply@anthropic.com>

coilyco-ops requested review from coilysiren

2026-08-15 21:59:45 +00:00