bc-dgx7.77: B7e-role — A relay role's whole operating contract, in one call

[mordam]

Adds bin/b7e-role: a relay role's whole operating contract in one call, closing the gap
five deluvia sessions each paid for by hand-slicing docs/STUDIO_CHARTER.md (726 lines),
APPROVAL_PIPELINE.md and a role file differently every time a relay opened or changed
roles — one of them reading a rule off trunk that a later ruling had already superseded
because its hand-cut slice happened to stop just short of the correction.

b7e-role -w <workspace> <role> prints the role's own charter block, the studio's one
law, the role file in full, its declared output template, and the legal --next targets
(lib/relay.js's rolesOf) — each annotated with a source path and line range so a disputed
sentence can be checked against the file it came from. --next alone answers the
question that cost four of the five sessions a rejected beadcause-relay --next call,
and --all lists the roster with each role's department and gate. Every path is read off
that workspace's own config.relays[name] entry (lib/relay.js) — the same definition
beadcause-relay and the relay brief already use — so deluvia is only the first caller,
not a hardcoded assumption.

lib/rolecontract.js does the markdown slicing as pure functions over already-read text.
The one non-obvious piece: clio's real output template is a worked example wrapped in a
fenced code block that is itself written in markdown headings (## Clio's Continuity Report, ### 1. Timeline) — a naive heading scanner reads the example as more of the
document and cuts the template off after two lines, so section-finding runs against a
fence-blanked copy of the text (same length, same line breaks) while every slice returned
is still cut from the original.

Wired per bc-dgx7.2: both bin registrations, Bash(b7e-role:*) on DEFAULT_TOOL_LIST
classified read in lib/grants.js, a README section, and test/role.mjs. The CLI half of
the test suite builds its own fixture workspace on disk with an isolated
BEADCAUSE_CONFIG_DIR, so it needs no deluvia checkout to run; a final block checks the
shipped deluvia default's roster shape (19 roles, 18 legal next-roles for clio) using
only --all/--next, which read no files at all, so that block holds on a machine that
has never heard of deluvia — the same guarantee test/relay.mjs's own shipped-config check
relies on.

Unsure of one thing: main moved under me while I worked (another branch added
b7e-rebaseline between the same two package.json bin entries b7e-role lands
between), so the registration lines will very likely collide on merge — a routine
take-both, per this repo's own notes on b7e-registry conflicts, but flagging it since I
haven't seen the other branch's actual diff.

Tests: node test/role.mjs (55/55 standalone), then the full local gate: bin/b7e-gate --jobs 3 — 478/478 suites passed (including test/role.mjs and test/lockfile.mjs).

Files changed — 8 files · +907 −0 · against `main`

test/role.mjs        +369 −0
bin/b7e-role         +314 −0
lib/rolecontract.js  +148 −0
README.md            +65 −0
lib/toolbelt.js      +8 −0
lib/grants.js        +1 −0
package-lock.json    +1 −0
package.json         +1 −0

---

Opened by a beadcause worker session on bc-dgx7.77 — b7e-role — A relay role's whole operating contract, in one call, instead of slicing a 726-line charter. It merges itself once the checks report; merging is what closes the bead. If this is still open, something stopped that, and the reason is on bc-dgx7.77 and in Adam's inbox.

bead: bc-dgx7.77

[mordam]

A beadcause worker opened this and does not merge its own work. It is on the merge queue as bc-ovby9.

[mordam]

The beadcause merge queue tried to merge this and could not: it has been tried 3 times and stopped at the same place each time. the branch still conflicts with main and no resolver is on it any more — the window that was opened on it ended without making it mergeable, so this one is yours to settle. Tried 3 times — that was the last. It is Adam's call now — see bc-ovby9.

[mordam]

Updated — 2 commits on worktree-b7e-role-dgx777 — 8 files, +981 −0 against main.

Follow-up to the same bead, same branch/PR: switches bin/b7e-role's directory default
from process.cwd() to lib/session.js's resolveSessionDir(cfg, ws), falling back to cwd.
Found via beadcause-memory notes a-checkout-dir-is-not-a-workspaces-dir, read right after
the first delivery — it warns that cfg.workspaces[name].dir is the bd tracker's directory,
never the code checkout, and that a new b7e-* command should use resolveSessionDir (the
pair ~25 existing call sites already share) rather than rely on an incidental cwd match.
Previously b7e-role -w deluvia clio only found the real files when run from inside the
deluvia checkout; now it finds them from anywhere, verified against the real checkout on
this Mac from inside the beadcause worktree itself. --dir still overrides outright, and
every previously-passing test is unchanged; one new test covers the no-dir/cwd-elsewhere
path.

Tests: node test/role.mjs (57/57 standalone, ~19s). Full gate (bin/b7e-gate --jobs 3) ran clean 478/478 before this follow-up commit; a second full-gate run afterward hit 52 unrelated timeouts under heavy concurrent Mac load (test/role.mjs included, as a 939s timeout) but passes cleanly standalone — the documented gate-reds-on-a-loaded-mac pattern, not a regression.

[mordam]

A beadcause worker opened this and does not merge its own work. It is on the merge queue as bc-72dnr.

[mordam]

The beadcause merge queue tried to merge this and could not: it has been tried 3 times and stopped at the same place each time. the branch still conflicts with main and no resolver is on it any more — the window that was opened on it ended without making it mergeable, so this one is yours to settle. Tried 3 times — that was the last. It is Adam's call now — see bc-72dnr.

[NeanderthalMan]

beadcause-resolver: stood down — worktree-b7e-role-dgx777 is already locked by another resolver on this same PR (reason "resolver pid 64046 #758", pid live in ps). Left the tree untouched and unlocked nothing; that session owns the merge.