@gobing-ai/superskill 0.1.2 → 0.1.4

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@gobing-ai/superskill",
-    "version": "0.1.2",
+    "version": "0.1.4",
     "description": "A manager for multi-agent skill, slash command, subagent, hook, MCP and etc.",
     "keywords": [
         "cli",
@@ -23,6 +23,7 @@
     "files": [
         "dist/",
         "templates/",
+        "rubrics/",
         "README.md"
     ],
     "engines": {
@@ -30,15 +31,16 @@
     },
     "scripts": {
         "start": "bun run src/index.ts",
-        "build": "bun build src/index.ts --outfile ../../dist/cli/index.js --target bun && bun run ../../scripts/postbuild.ts ../../dist/cli/index.js",
-        "prepublishOnly": "bun run build && mkdir -p dist && cp ../../dist/cli/index.js dist/index.js",
+        "build": "bun build src/index.ts --outfile dist/index.js --target bun && bun run ../../scripts/builder.ts postbuild dist/index.js && rm -rf templates rubrics && cp -r src/templates templates && cp -r ../../packages/core/src/rubrics rubrics",
+        "prepublishOnly": "bun run build && cp ../../README.md README.md",
         "test": "NODE_ENV=test bun test --reporter=dots",
         "typecheck": "tsc --noEmit"
     },
     "dependencies": {
-        "@gobing-ai/ts-ai-runner": "^0.3.19",
-        "@gobing-ai/ts-db": "^0.3.19",
-        "@gobing-ai/ts-utils": "^0.3.19",
+        "@gobing-ai/superskill-core": "workspace:*",
+        "@gobing-ai/ts-ai-runner": "^0.3.21",
+        "@gobing-ai/ts-db": "^0.3.21",
+        "@gobing-ai/ts-utils": "^0.3.21",
         "commander": "^14.0.0",
         "drizzle-orm": "^0.44.0",
         "drizzle-zod": "^0.7.0",

package/rubrics/agent.yaml

@@ -0,0 +1,52 @@
+version: 1
+type: agent
+dimensions:
+  - name: completeness
+    weight: 0.20
+    criterion: >
+      Does the agent cover its stated scope end-to-end? Penalize missing workflow
+      phases, undocumented edge cases, and gaps between the stated trigger and
+      the actual execution path.
+    anchors:
+      excellent: "Full workflow from intake to delivery; every phase has explicit steps and exit conditions."
+      poor: "Scope claims coverage but execution path skips phases or leaves edge cases unhandled."
+
+  - name: role-clarity
+    weight: 0.25
+    criterion: >
+      Does the body define a specific, non-generic persona with a clear scope?
+      Penalize "helpful assistant" framing and vague responsibilities. The role
+      name must convey expertise and boundary, not just willingness to help.
+    anchors:
+      excellent: "Named specialist with explicit expertise boundary, e.g. 'SECU security reviewer for auth flows'."
+      poor: "Generic 'helpful assistant that assists with tasks' — no specialization, no boundary."
+
+  - name: tool-selection
+    weight: 0.20
+    criterion: >
+      Are tool choices appropriate and justified? Each tool reference should
+      match the task's needs; penalize missing tools that the workflow requires
+      or tools listed but never invoked in the execution path.
+    anchors:
+      excellent: "Every tool is matched to a specific step; no missing or orphaned tool references."
+      poor: "Tools listed generically or mismatched to the workflow's actual needs."
+
+  - name: skill-linkage
+    weight: 0.20
+    criterion: >
+      Does the agent link to the right skills at the right time? Skill
+      references should appear where the workflow delegates, not as a bulk
+      list. Penalize skills referenced but never reached in execution.
+    anchors:
+      excellent: "Skills are linked at the exact delegation point in the workflow; each link is reachable."
+      poor: "Skills listed in a block with no workflow context on when to invoke each."
+
+  - name: model-fit
+    weight: 0.15
+    criterion: >
+      Is the recommended model tier appropriate for the agent's cognitive
+      load? A routing/lookup agent should not mandate a slow model; a deep
+      reasoning agent should not default to a fast model.
+    anchors:
+      excellent: "Model tier matches cognitive demand; routing uses fast, synthesis uses slow, with rationale."
+      poor: "Single model tier with no justification, or mismatched to the task's reasoning depth."

package/rubrics/command.yaml

@@ -0,0 +1,52 @@
+version: 1
+type: command
+dimensions:
+  - name: completeness
+    weight: 0.25
+    criterion: >
+      Does the command cover its stated function end-to-end? Penalize missing
+      argument validation, undocumented flags, and incomplete output. A
+      complete command handles the happy path and the documented error paths.
+    anchors:
+      excellent: "All arguments validated, all flags documented, output covers success and error paths."
+      poor: "Happy path only; argument validation and error paths undocumented."
+
+  - name: clarity
+    weight: 0.25
+    criterion: >
+      Is the command's purpose and usage unambiguous? The description should
+      tell the user what it does and when to use it. Penalize descriptions
+      that restate the command name without adding information.
+    anchors:
+      excellent: "Description states what the command does, when to use it, and what the output means."
+      poor: "Description restates the command name; user cannot tell when to use it."
+
+  - name: argument-hints
+    weight: 0.20
+    criterion: >
+      Are argument hints present and accurate? Each argument should have a
+      hint that helps the user supply the right value. Penalize missing hints,
+      hints that restate the argument name, and hints that mislead.
+    anchors:
+      excellent: "Every argument has a hint that guides the user to the correct value."
+      poor: "Hints missing, or restate the argument name without guidance."
+
+  - name: tool-references
+    weight: 0.15
+    criterion: >
+      Are tool references (if any) correct and reachable? Penalize commands
+      that reference tools the CLI does not expose, or tools that are exposed
+      but never documented in the command's usage.
+    anchors:
+      excellent: "Tool references match exposed CLI verbs; no orphaned or undocumented references."
+      poor: "Tool references are stale, orphaned, or undocumented."
+
+  - name: slash-syntax
+    weight: 0.15
+    criterion: >
+      Is the slash command syntax correct and consistent? Penalize syntax
+      that diverges from the platform convention, inconsistent argument
+      ordering, and missing required markers.
+    anchors:
+      excellent: "Syntax follows platform convention; argument ordering is consistent across sibling commands."
+      poor: "Syntax diverges from convention or is inconsistent with sibling commands."

package/rubrics/hook.yaml

@@ -0,0 +1,42 @@
+version: 1
+type: hook
+dimensions:
+  - name: correctness
+    weight: 0.30
+    criterion: >
+      Does the hook's logic produce the intended effect? Penalize hooks that
+      fire on the wrong event, mutate state the spec says they shouldn't, or
+      exit with the wrong code on a documented path.
+    anchors:
+      excellent: "Logic matches the spec on every documented path; correct event, mutation, and exit code."
+      poor: "Logic diverges from spec on at least one documented path."
+
+  - name: event-coverage
+    weight: 0.30
+    criterion: >
+      Does the hook handle all events it claims to? Penalize hooks that
+      register for an event but silently no-op on it, or that miss an event
+      the spec requires. Coverage means the hook acts on every registered event.
+    anchors:
+      excellent: "Every registered event has a real handler; no silent no-ops; no missing required events."
+      poor: "Registered events include silent no-ops, or required events are missing."
+
+  - name: safety
+    weight: 0.25
+    criterion: >
+      Does the hook avoid destructive side effects? Penalize hooks that
+      force-push, delete without confirmation, bypass gates, or modify files
+      outside their stated scope. Safety is about blast radius, not just correctness.
+    anchors:
+      excellent: "No destructive side effect without an explicit guard; blast radius matches stated scope."
+      poor: "Destructive side effects (force-push, delete, bypass) without guards."
+
+  - name: pattern-match-quality
+    weight: 0.15
+    criterion: >
+      Are the matchers precise? Penalize matchers that over-match (fire on
+      unrelated commands) or under-match (miss the intended case). A precise
+      matcher fires on exactly the intended inputs.
+    anchors:
+      excellent: "Matchers fire on exactly the intended inputs; no over- or under-matching."
+      poor: "Matchers over-match (fire on unrelated commands) or under-match (miss intended case)."

package/rubrics/magent.yaml

@@ -0,0 +1,52 @@
+version: 1
+type: magent
+dimensions:
+  - name: completeness
+    weight: 0.25
+    criterion: >
+      Does the main-agent config cover its stated scope end-to-end? Penalize
+      missing routing rules, undocumented fallbacks, and gaps between the
+      stated trigger and the actual execution path.
+    anchors:
+      excellent: "Routing, fallbacks, and execution path are fully explicit; no implicit steps."
+      poor: "Scope claims coverage but routing or fallbacks are implicit."
+
+  - name: platform-coverage
+    weight: 0.25
+    criterion: >
+      Does the config address all platforms it claims to support? Penalize
+      platforms listed in scope but with no corresponding rules, and rules
+      that reference capabilities a listed platform lacks.
+    anchors:
+      excellent: "Every listed platform has corresponding rules; rules respect each platform's capabilities."
+      poor: "Platforms listed without rules, or rules reference capabilities a platform lacks."
+
+  - name: conciseness
+    weight: 0.15
+    criterion: >
+      Is the config as short as possible while remaining complete? Penalize
+      redundant rule restating, boilerplate that adds no information, and
+      rules that could be merged without loss.
+    anchors:
+      excellent: "Minimal length with no redundancy; every rule carries unique information."
+      poor: "Redundant rules and boilerplate; rules could be merged without loss."
+
+  - name: tone-consistency
+    weight: 0.20
+    criterion: >
+      Is the agent's tone consistent across the config? Penalize mixed
+      registers (formal header + casual body), inconsistent second/third
+      person, and persona drift between sections.
+    anchors:
+      excellent: "Single register and person throughout; persona is stable across all sections."
+      poor: "Mixed registers, inconsistent person, or persona drift between sections."
+
+  - name: safety
+    weight: 0.15
+    criterion: >
+      Does the config avoid dangerous defaults? Penalize configs that grant
+      blanket permissions, disable safety gates by default, or route
+      destructive actions without an explicit guard.
+    anchors:
+      excellent: "Least-privilege defaults; safety gates enabled; destructive actions guarded."
+      poor: "Blanket permissions or disabled safety gates without justification."

package/rubrics/skill.yaml

@@ -0,0 +1,53 @@
+version: 1
+type: skill
+dimensions:
+  - name: completeness
+    weight: 0.25
+    criterion: >
+      Does the skill cover its stated purpose end-to-end? Penalize missing
+      steps, undocumented preconditions, and gaps between the trigger and the
+      delivered outcome. A complete skill leaves no implicit step.
+    anchors:
+      excellent: "Every phase from trigger to outcome is explicit; preconditions and exit conditions stated."
+      poor: "Purpose claims coverage but steps are implicit or preconditions unstated."
+
+  - name: clarity
+    weight: 0.25
+    criterion: >
+      Is the skill's instruction unambiguous to a fresh agent? Penalize
+      vague verbs ('handle', 'process'), undefined terms, and instructions
+      that require external context the skill does not provide.
+    anchors:
+      excellent: "Concrete verbs, defined terms, self-contained instructions an agent can execute without guessing."
+      poor: "Vague verbs and undefined terms; agent must guess the intended behavior."
+
+  - name: trigger-accuracy
+    weight: 0.20
+    criterion: >
+      Does the skill fire on the right inputs and not on adjacent ones? The
+      description must precisely scope when the skill activates. Penalize
+      triggers that are too broad (fires on unrelated requests) or too narrow
+      (misses the intended case).
+    anchors:
+      excellent: "Trigger description scopes activation precisely; no overlap with adjacent skills."
+      poor: "Trigger is broad enough to fire on unrelated requests or narrow enough to miss the intended case."
+
+  - name: anti-hallucination
+    weight: 0.15
+    criterion: >
+      Does the skill prevent fabrication? Penalize instructions that invite
+      guessing (e.g. 'infer the answer'), missing verification gates, and
+      absence of source-grounding requirements where facts are involved.
+    anchors:
+      excellent: "Explicit verification gates and source-grounding; no instruction invites guessing."
+      poor: "Instructions invite inference without verification; no grounding requirement."
+
+  - name: conciseness
+    weight: 0.15
+    criterion: >
+      Is the skill as short as possible while remaining complete? Penalize
+      redundant restating, boilerplate that adds no information, and steps
+      that could be merged without loss.
+    anchors:
+      excellent: "Minimal length with no redundancy; every line carries unique information."
+      poor: "Redundant restating and boilerplate; steps could be merged without loss."

package/templates/agent/default.md

@@ -0,0 +1,8 @@
+---
+name: <!-- NAME -->
+description: <!-- DESCRIPTION -->
+tools: []
+model: sonnet
+---
+
+<!-- TODO: agent system prompt and configuration -->

package/templates/command/default.md

@@ -0,0 +1,16 @@
+---
+name: <!-- NAME -->
+description: <!-- DESCRIPTION -->
+arguments: []
+target: <!-- TARGET -->
+---
+
+# <!-- NAME -->
+
+## Usage
+
+```text
+/<!-- NAME -->
+```
+
+<!-- TODO: command body -->

package/templates/hook/default.md

@@ -0,0 +1,8 @@
+---
+name: <!-- NAME -->
+description: <!-- DESCRIPTION -->
+event: PreToolUse
+enabled: true
+---
+
+<!-- TODO: hook script or matcher -->

package/templates/magent/default.md

@@ -0,0 +1,22 @@
+---
+name: <!-- NAME -->
+description: <!-- DESCRIPTION -->
+platforms:
+  - claude
+---
+
+## IDENTITY
+
+<!-- TODO: who the agent is at a glance -->
+
+## SOUL
+
+<!-- TODO: tone contract, communication examples, decision-making style -->
+
+## AGENTS
+
+<!-- TODO: operations: routing, tools, workflow, safety, verification -->
+
+## USER
+
+<!-- TODO: operator profile and preferences -->

package/templates/skill/default.md

@@ -0,0 +1,8 @@
+---
+name: <!-- NAME -->
+description: <!-- DESCRIPTION -->
+---
+
+# <!-- NAME -->
+
+<!-- TODO: skill body -->