Resolve DQ04: a Plan's identity is its origin - #2

Replace the operator-chosen catalog directory name with a derived PlanRef: the
content hash of a Plan's origin — the single predecessor-less version. Identity
is derived, never chosen, never minted, and encodes no location.

- catalog: derive_planref walks predecessor imports to the origin and hashes it;
  an origin is its own PlanRef (its version-id and the PlanRef are the same hash).
  load_plan rejects a version whose derived origin != the dir it is filed under,
  on the same terms as a content-hash-vs-filename mismatch — never reinterpreted.
- cli/cmd: start names no plan and scaffolds into a drafts/ staging area with a
  blank (required) goal; commit derives the PlanRef and files under it — --plan
  and <plan> args removed. Reads address a Plan by PlanRef (the dir), a unique
  hash prefix, or, as a nicety, an unambiguous goal.
- model: a non-empty goal is required on every version (CMP.DM-R12), checked on
  the evaluated value so revisions inheriting their goal pass.
- display: status and history lead with the goal (the human handle), not the hash.
- examples: re-file the three plans under their origin hashes (bytes unchanged);
  READMEs refer to each plan by its goal and note the PlanRef is the origin hash.
- tests: cover origin-files-under-own-hash, revision-shares-PlanRef, misfiled
  rejection, empty-goal refusal, and goal display; rewrite the e2e + cross-plan
  suites for the derived-identity, no-name-arg flow.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Complete decision 0017's "goal is the human handle, surfaced wherever a Plan is
referenced": the progress and evidence confirmation lines printed the raw
PlanRef; show the goal instead (the version is still cited by its hash). Add an
acceptance test asserting both commands render the goal.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…mple

The DQ04 edit added CMP.DM-R11/R12 in the Identity section, colliding with
the pre-existing Progress/acceptance R11/R12 — a duplicate-id bug that
axe vrs check did not catch. Renumber the new pair to R17 (Plan identity)
and R18 (goal), leaving the older Progress/Acceptance ids and the
07-evals CMP.DM-R12 citation valid. Update the impl comments that cited
the new ids.

Also: decision 0017 cited CMP.DM-R08 (Step identity = declared name) as
forbidding location-as-identity; the no-location property is the PlanRef
ontology entry — corrected. And the 06-api cross-plan example showed a
pl_-named dir; under 0017 a plan segment is the origin hash, so the
example now uses one and refers to the other plan by its goal.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
EOF

The *Ref terms were coined for the minted-reference design, where identities
were opaque tokens you passed around. That design is gone: a Step's identity
is its declared export name (0012) and a Plan's is its origin hash (0017) —
intrinsic identities, not reference tokens.

- StepRef is dropped as a standalone term. There is no ref artifact: a Step's
  identity is its name, and depending on a Step names the declaration (a
  language reference). Folded into the Step ontology entry.
- PlanRef -> PlanId: it is an identity (a derived hash), not a pointer to
  elsewhere; a cross-Plan reference imports the other version.

Propagated across the current layer (specs, api, cli, examples, code incl.
derive_planref -> derive_planid). The superseded decision 0004 keeps the old
terms as a historical record; the ontology's _Avoid_ lists PlanRef/StepRef so
a reader of old records maps them. 130 lib + 15 acceptance, clippy/fmt/vrs green.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
EOF

The 22-term list was flat. Group under six headings so the structure is
legible without reading every entry: the tool; intent (Plan, PlanId,
Version, Revision, Rationale, Step); lineage and its states (Head,
Divergence, Reconciliation, Orphan, Unresolved); reading, storage, and
replication; execution record; change surface and composition. Term text
unchanged.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…name predecessor→parent

Interview-resolved ontology-refinement pass:

- Surface cluster: drop Plan Surface, Receipt, Observation as terms. Anchor on
  Commit as the single write act; the write surface has two acts (Commit →
  Plan Version, Progress Event → append), not one Mutation umbrella. The
  Plan Surface invariant survives as a property; Receipt's role is the version
  hash itself; Observation's guarantee relocates to 05-integrations.
- Add Goal, Acceptance, Evidence as terms. Strike phantom Gate — the acceptance
  predicate is the gate.
- Rename predecessor → parent (aligns with the `parents` the impl already emits).
- Split lineage terms: "Lineage shapes" (Head, Divergence, Reconciliation) vs
  "Incomplete replication" (Orphan, Unresolved).
- Rationale keeps its term, bridged explicitly to the `why` field.

Mutation→Commit and Receipt→Plan Version propagated across specs/requirements
and the governing 0001 decision; CMP.SURF-R03 rescoped to the two-act model.
Decisions and frozen example modules keep their period vocabulary.

axe vrs check --profile strict: ok

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

- Rename predecessor → parent across src and tests (helper fns, error strings,
  doc-comments, test names). The JSON already emitted `parents`; this closes the
  gap between the code's internal words and what it produces.
- Rename receipt_json → commit_result_json: the output has no `receipt` key, and
  the ontology no longer defines Receipt — the returned value is the Plan Version.
- Rewrite the readiness "On gates" note: the spec no longer says "dependencies
  and gates"; the ontology states the acceptance predicate *is* the gate.

Frozen example modules under catalog/plans/*/versions/*.ts are untouched — they
keep their period vocabulary, and the acceptance hash oracle
(example_filenames_reproduce_from_source_bytes) confirms none drifted.

cargo test: 15 passed. clippy + fmt clean.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>