arcane-cli 0.25.0 → 0.26.0

package/dist/assets/.claude/commands/spell-scry.md

@@ -0,0 +1,13 @@
+---
+description: Use PROACTIVELY when naming anything new (a spell, a field, a product, an internal concept) before committing to it — clears a candidate name against both the outside world and this repository itself.
+---
+
+# Scry
+
+Invoke the Arcane `spell-scry` spell workflow.
+
+See the full prompt at `.github/prompts/spell-scry.prompt.md` for the complete workflow definition.
+
+---
+
+@.github/prompts/spell-scry.prompt.md

package/dist/assets/.github/prompts/spell-scry.prompt.md

@@ -0,0 +1,123 @@
+---
+name: Spell — Scry
+description: Clear a candidate name before it ships — four outward checks (who coined it, estate still trading, same-audience giants, first-association salience) plus a mandatory inward repo-local collision pass. Use when naming anything new (a spell, a field, a product, an internal concept) before committing to it.
+argument-hint: '<candidate name or term>'
+agent: agent
+---
+
+## Executive Summary
+
+- This spell clears a candidate name before it ships, checking both the outside world and this
+  repository itself.
+- The inward pass runs first and is mandatory, not optional — a real naming decision (ARC-028's
+  `workspace`) was nearly settled on a clean external read alone, until a single `grep -ri` found the
+  word already load-bearing inside the repository in two other senses.
+- Both passes classify hits with the same taxonomy (same-space / adjacent / out-of-space) and combine
+  into one verdict: pass / pass-with-disclosure / kill.
+- `spell ward` is the opposite direction of this same concern — it finds what leaked **in**; this spell
+  clears what's about to go **out**. They do not call each other; ward's denylist scanning and scry's
+  single-candidate classification are different operations on the same underlying grep primitive.
+
+---
+
+Clear the candidate name given in the prompt argument.
+
+## Step 1 — Resolve the candidate
+
+Take the term verbatim from the argument. If none was given, ask for it — do not guess a name to check.
+
+## Step 2 — Inward pass (mandatory, run first)
+
+Grep this repository for the candidate term before doing any outward research:
+
+- **Prose**: markdown, comments, documentation.
+- **Identifiers**: variable, function, type, and class names — weight these **above** prose, since
+  consumers already depend on code identifiers in a way they never depend on a sentence mentioning a
+  word.
+- **Config keys and schema fields**: `.arcane.json` fields, YAML/JSON config keys, any validated schema.
+
+Use word-boundary matching (the same principle `spell ward`'s denylist scan uses, expressed here as a
+`grep -w` / IDE-search equivalent over the codebase) — a bare substring match on a short candidate term
+will drown in false positives from unrelated words that merely contain it.
+
+**Classify each internal hit with the same taxonomy the outward checks use below:**
+
+- **Same-space**: the term is already load-bearing for a closely related concept (a config field, a
+  directory name, a class doing adjacent work) — a genuine internal collision.
+- **Adjacent**: the term appears near the concept but for something distinguishable enough that a
+  reader would not confuse the two once told them apart.
+- **Out-of-space**: incidental — a comment, an unrelated variable, coincidental substring overlap.
+
+**Why this runs first:** a same-space internal collision is often cheaper to find than the outward
+research below, and finding one early can save the effort of a full outward pass on a name that was
+already going to be killed. This is not a formality — the ARC-028 `workspace` incident is the concrete
+case: a clean external read had nearly settled the decision before a single `grep -ri` found the word
+already meaning something else in two other places inside this exact repository.
+
+## Step 3 — Outward pass: the four checks
+
+Research the candidate term against the outside world:
+
+1. **Who coined it** — search for the term's origin and earliest known usage in the relevant space
+   (software, apps, AI, games — whichever applies to this candidate).
+2. **Is an estate/trademark holder still trading** — distinguish an actively-used, enforced brand from
+   an abandoned or long-defunct one. A defunct mark carries much less collision risk than a live one.
+3. **Same-audience giants** — do major, well-known players who share this candidate's actual target
+   audience already use this name saliently? A collision with a giant outside the target audience
+   matters far less than one inside it.
+4. **First-association salience per market** — for someone in the target market, what does this term
+   bring to mind *first*? If the first association is something else entirely, that is a real cost even
+   without a formal trademark conflict.
+
+Classify each outward finding with the **same** same-space / adjacent / out-of-space taxonomy as the
+inward pass — never a separate scheme. Cite sources for every outward claim; do not assert a name's
+history or salience from unverified general knowledge alone.
+
+## Step 4 — Combine into one verdict
+
+A name that clears outward but collides inward is not a clean pass, and the reverse is equally true —
+combine both passes into a single verdict, not two separate reports:
+
+- **pass** — no same-space finding, inward or outward; adjacent/out-of-space findings, if any, are
+  noted but do not block.
+- **pass-with-disclosure** — a same-space finding exists but is distant, low-traffic, or otherwise minor
+  enough that proceeding is reasonable with the collision stated explicitly, not discovered later.
+- **kill** — a genuine, blocking same-space collision, inward or outward. Do not soften this to
+  pass-with-disclosure merely because the outward research took effort to produce.
+
+## Step 5 — Report
+
+```markdown
+## Scry verdict: <candidate> — PASS | PASS-WITH-DISCLOSURE | KILL
+
+### Inward pass
+| Hit | Location | Classification |
+|---|---|---|
+| <term usage> | <file:line> | same-space / adjacent / out-of-space |
+
+### Outward pass
+| Check | Finding | Classification | Source |
+|---|---|---|---|
+| Who coined it | ... | ... | <url or citation> |
+| Estate still trading | ... | ... | <url or citation> |
+| Same-audience giants | ... | ... | <url or citation> |
+| First-association salience | ... | ... | <url or citation> |
+
+### Verdict rationale
+<why this combination of inward + outward findings produced this verdict>
+```
+
+## Related Spells
+
+- `spell ward` — the opposite direction: finds third-party identifiers that already leaked into the
+  repo, rather than clearing a candidate before it ships out.
+
+## Rules
+
+- **Never skip the inward pass.** It is not optional, and it runs before the outward pass, not after.
+- **Never assert an outward finding without a source.** "Based on general knowledge" is not a citation.
+- **Never use a separate classification scheme for inward vs. outward findings.** Same-space / adjacent
+  / out-of-space applies to both, so a reader can compare them directly.
+- **Never round pass-with-disclosure up to pass, or down to kill, to avoid stating a nuanced verdict.**
+- This spell classifies and reports. It does not rename, alias, or otherwise resolve a collision it
+  finds — that decision belongs to whoever asked for the name cleared.

package/dist/index.js

@@ -551,7 +551,9 @@ var COMPONENTS = [
       ".github/prompts/spell-generate-bot-icons.prompt.md",
       ".claude/commands/spell-generate-bot-icons.md",
       ".github/prompts/spell-make-discoverable.prompt.md",
-      ".claude/commands/spell-make-discoverable.md"
+      ".claude/commands/spell-make-discoverable.md",
+      ".github/prompts/spell-scry.prompt.md",
+      ".claude/commands/spell-scry.md"
     ]
   },
   {