OODA: lineage and a sidepeek for every stage (PR 3 of 5) - #201

…p card opens

An observation, a recurring signal and the learning record had no sidepeek,
and a Loop card opened only from its timestamp. Now each opens its own: an
observation's says where it came from first (its computed provenance, in
plain words) and where it sits in the loop (not oriented yet, n of the 3 a
signal takes, explained, ruled noise, ignored, set aside, or carried on to
the furthest step), which the kit computes once (placeOf) for the card and
the peek alike. A signal opens by its key, / and all, with its observations,
how long it has waited and what would explain it. One delegated handler
opens any [data-open] card, leaving a link, a control, a modified click, a
pinned thread layer and a selection alone; the incidents' own handlers go,
so nothing fires twice. o- and q- ids in text link to their sidepeeks.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

# Conflicts:
#	docs/devteam/scripts/devteam.mjs

…ent's fingerprint

ooda-signals made explained() the one definition of what explains an
observation, keys included: a noise ruling that names a signal's key covers
its later observations, and an open incident holds its fingerprints. A card
built on a second definition would say "not oriented yet" for a spike the
incident already holds; now it says the incident explains it, or that it
was ruled noise.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Each OODA sidepeek (observation, signal, lesson, problem, decision, act,
learning) now shows how the record came about and what came of it.
renderMermaid's initialize, render and sanitize are factored into
drawMermaid(text) and shared. The list, grouped by stage with every record
a link, is the first paint and what stays when Mermaid can't load; the graph,
once drawn, has each node wrapped in a link to its sidepeek and is cached by
its nodes' ids, so a refresh that changes none of them leaves the drawer as
it is. Mermaid 11.4.1's sanitized SVG carries no data-id on node groups, so
a node is found by data-id or by its flowchart-n<i>-<counter> id. The raw
Mermaid source is never shown.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

A recurring signal can sit unexplained for days with nothing saying so.
dashboard-data's ooda.debt.signals lists the ones unexplained for more than
a day, oldest first, read off the same explained() Orient uses, and the
Orient tab shows them above the recurring signals, each card opening its
signal. Resolved incidents nobody learned from join it in PR 4.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

The box's operator reviews how each repo's loop runs, and the lineage is
in peek (--json's lineage, --mermaid). peek only reads: the operator's
read-only test now runs peek for a lesson (--json and --mermaid), a
problem, a goal and a sensor, and the files and git status are what they
were.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

supervisor.md §3 step 4 says what orientation debt is and where it shows,
and how to read a record's lineage (peek --json) or paste it into a retro or
a PR body (peek --mermaid).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

main's golden was recorded before the OODA PRs. dashboard-data's
observations now carry stage and provenance (PR 2) and place (PR 3), ooda
carries debt, and peek ticket says stage null. Each is a key added beside
what was there; no value a single-repo project printed before is changed or
gone, so the golden is recorded again rather than the views narrowed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

ooda-signals' explained() now maps a signal key to until when it is
explained, read with covered(): a noise ruling naming a key covers that
signal's observations up to the ruling. Where an observation sits reads it
the same way (one after a ruling is not oriented yet, not ruled noise), and
orientation debt follows through orientClusters. supervisor.md keeps both
sides' sentences.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

- linkIndex: an incident holds each observation on its fingerprints from its
  opening to its resolution ('held'), so a later spike's sidepeek has the
  incident in its lineage and its place follows it on
- one edge a pair: a second field citing it joins its why ("evidence, asks about")
- inbox ("Asked you") and measurement ("Measured") stages drawn in their own
  groups, with a stroke each, in the CLI's Mermaid and the page's list
- peek signal <unknown> --mermaid dies as --json does
- the page: Orientation debt and Recurring signals cards have a link inside;
  the lineage graph is kept by node id, stage and label, and a failed render
  is tried again on the next open; a signal's "Recurring" is blank while loading
- tests: held edges, the hop bound, joined whys, group edges to two neighbours
- spec §4: the ad-hoc indexes stay as they are for now

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…earch and peek

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…graph cache

Review fixes from the upstream loop: an approved_by edge reads approved only
when the approval checks out (a decision with no authority gap, an act
carrying one out, an act resting on the owner's accepted rule or an answered
question); anything else cites it, so a supervisor's lesson never reads as
the owner approving. The help names --mermaid and the new kinds, and the
page keeps the last 50 drawn graphs.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>