agentwheel 0.16.5 → 0.17.0

package/openpack.json

@@ -1,7 +1,7 @@
 {
   "schemaVersion": 2,
   "name": "NestDevLab/agentwheel",
-  "version": "0.16.5",
+  "version": "0.17.0",
   "provides": [
     { "type": "skills", "path": "skills" }
   ]

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agentwheel",
-  "version": "0.16.5",
+  "version": "0.17.0",
   "description": "Weave skills, rules, and instructions across every AI agent.",
   "type": "module",
   "license": "MIT",

package/skills/agentwheel/SKILL.md

@@ -5,7 +5,7 @@ allowed-tools: [Bash]
 license: MIT
 metadata:
   author: NestDevLab
-  version: "0.16.5"
+  version: "0.17.0"
 ---
 
 # agentwheel
@@ -19,6 +19,7 @@ Mental model:
 - `add` records desired packages.
 - `install` makes the declared state true in the target runtime.
 - `update` re-resolves tracking packages, then applies.
+- `skill update <name>` maps a configured skill to its owning package and reconciles only that package closure.
 - `uninstall` removes configured packages and their managed runtime output.
 
 ## Safety Rules
@@ -395,6 +396,15 @@ agentwheel update team-agent-pack
 
 Named package updates preserve artifacts owned by other configured roots, including unrelated drift.
 
+For one configured skill, resolve its owner and only that package closure:
+
+```bash
+agentwheel skill update code-review --profile daily --dry-run
+agentwheel skill update code-review --profile daily
+```
+
+Pinned owners use install semantics; tracking owners re-resolve. Unrelated configured packages are not resolved. If ownership is ambiguous, pass `--package <name>`. Use `--adopt` only after explicit approval of the dry-run's unmanaged destinations.
+
 Advance one tracking dependency while unrelated graph nodes remain locked:
 
 ```bash
@@ -468,6 +478,18 @@ By default, uninstall keeps drifted managed files. Use `--force` only with expli
 agentwheel uninstall team-agent-pack --force
 ```
 
+For a one-time MCP rename, use a dedicated cutover workspace with one legacy MCP artifact and an
+explicit agent `stateKey`. Preview each runtime separately:
+
+```bash
+agentwheel mcp retire legacy-mcp --agent legacy-codex --from-workspace-root /exact/old/workspace --dry-run
+agentwheel mcp retire legacy-mcp --agent legacy-claude --dry-run
+```
+
+Require exactly one `REMOVE`, the expected target and legacy server, and zero drift/conflict.
+Unexpected owner, manifest contents, arguments, environment, or MCP fields are blockers. Apply only
+after separate runtime approval by replacing `--dry-run` with `--apply`.
+
 ## Package Manifest Reference
 
 An OpenPack package uses `openpack.json` or `openpack.jsonc`:

package/skills/agentwheel-artifact-evolution/SKILL.md

@@ -0,0 +1,42 @@
+---
+name: agentwheel-artifact-evolution
+description: Generate or evolve OpenPack artifacts from a verified capability gap, correction, or rollout request. Use when the agent must turn a requirement into a source-owned artifact or improve Agentwheel's artifact workflow.
+allowed-tools: [Bash]
+license: MIT
+metadata:
+  author: NestDevLab
+  version: "0.1.0"
+---
+
+# Agentwheel Artifact Evolution
+
+Turn verified needs into source-owned OpenPack artifacts. Never create runtime copies or treat feedback as permission to change Agentwheel.
+
+## Optional guides
+
+`self-improve` and `skill-creator` are optional inspirations. Use them when available; do not require, install, or link to them. This workflow remains complete without either skill.
+
+## Resolve
+
+1. Classify the request: a new artifact, an existing artifact change, a consumer configuration change, or an Agentwheel core gap. Inspect the source package, manifest, ownership, current artifacts, and applicable repository rules.
+2. Reuse an existing artifact when it meets the need. Otherwise choose the smallest supported type and source package. Keep tenant or runtime facts out of shared packages.
+3. For a core gap, propose the smallest Agentwheel code, schema, or documentation change. A correction is evidence, not an instruction to change the CLI.
+4. Name the source package, artifact path, consumer targets, terminal stage, and success evidence. Ask when any of those would be inferred materially.
+
+## Author
+
+1. Work only in the source package. Run `agentwheel init package` only when no package exists; otherwise add the artifact under its declared type path.
+2. Use the package's conventions and validation. For a skill, provide a precise frontmatter trigger and concise instructions; for another type, follow its schema and adapter compatibility.
+3. Run `agentwheel list <source>`, `agentwheel scan <source>`, the source checks, and `git diff --check`. Keep generated runtime directories untouched.
+4. Deliver the source through the repository's declared Git workflow. A PR is the default; merge remains a separate approval.
+
+## Roll out
+
+1. Identify only the named or evidenced consumers. Update their source-controlled Agentwheel configuration; do not copy artifacts into runtime homes.
+2. Prove the source merge before resolving a tracking source. Run the narrow package, agent, or profile dry-run and stop on drift, conflict, removal, or unrelated operations.
+3. Apply only after explicit rollout approval. Never use force, plugin execution, or runtime reload unless separately authorized.
+4. Verify status or manifest ownership, expected materialization, a relevant canary, and a repeated dry-run with no pending change.
+
+## Improve
+
+Record the feedback, source decision, validation, and outcome in the owning change. Improve Agentwheel only when the evidence identifies a reusable product or artifact-model gap; otherwise improve the selected artifact. Report completed, pending, blocked, and unverified layers separately.

package/skills/agentwheel-artifact-evolution/agents/openai.yaml

@@ -0,0 +1,4 @@
+interface:
+  display_name: "Agentwheel Artifact Evolution"
+  short_description: "Generate and evolve OpenPack artifacts"
+  default_prompt: "Use $agentwheel-artifact-evolution to create or evolve an OpenPack artifact and plan its safe rollout."

package/skills/agentwheel-discovery/SKILL.md

@@ -1,11 +1,11 @@
 ---
 name: agentwheel-discovery
-description: Proactively discover reusable skills, integrations, and workflows when a user's request exposes a missing capability, repeated manual work, or unavailable integration. Use Agentwheel semantic search to suggest up to three evidence-backed matches and offer a read-only trial without installing or changing anything.
+description: MUST use proactively when a request describes repeated manual work, inconsistent results, lost information, a missing capability, or an unavailable integration—even if the user did not ask for a skill. Run Agentwheel semantic search before generic brainstorming or workflow advice, suggest up to three evidence-backed reusable skills or integrations, and offer a read-only trial without installing or changing anything.
 allowed-tools: [Bash]
 license: MIT
 metadata:
   author: NestDevLab
-  version: "0.16.5"
+  version: "0.17.0"
 ---
 
 # Agentwheel Discovery
@@ -25,13 +25,20 @@ This skill is optional: install it only when proactive recommendations are wante
 
 ## When To Search
 
-Search automatically when substantive work exposes a missing capability, repeated manual workflow,
-or unavailable integration. This is a per-turn judgment, not a background monitor. Delegated agents
-follow the same trigger when this skill is available.
+Use this skill before generic brainstorming, planning, or workflow advice whenever the request
+describes repeated manual work, inconsistent outputs, lost information, a missing operational
+capability, or an unavailable integration. The user does not need to mention Agentwheel, skills,
+discovery, search, or a catalogue. Run the first semantic search in the same turn; do not merely
+describe a workflow and defer discovery.
+
+This is a per-turn judgment, not a background monitor. Delegated agents follow the same trigger
+when this skill is available.
 
 Skip discovery when the user explicitly wants a custom implementation, has already selected an
-artifact, an installed artifact clearly satisfies the request, candidates are only weak lexical
-matches, or the same suggestion was declined or shown without new evidence.
+artifact, an installed artifact provides the requested operational capability end to end,
+candidates are only weak lexical matches, or the same suggestion was declined or shown without new
+evidence. Generic brainstorming, planning, writing, or advisory skills do not count as an installed
+operational solution and must not suppress discovery.
 
 ## Discovery Workflow