chore(plugins): regenerate plugin manifests

Reviewed-on: #9

.claude-plugin/marketplace.json

@@ -8,6 +8,10 @@
   "plugins": [
     {
       "name": "caveman",
+      "description": "Ultra-compressed communication mode from JuliusBrussee/caveman \u2014 cuts output tokens while keeping full technical accuracy. Installed straight from the upstream repo, not vendored here.",
+      "author": {
+        "name": "Rootiest"
+      },
       "source": {
         "source": "github",
         "repo": "JuliusBrussee/caveman"
@@ -15,30 +19,58 @@
     },
     {
       "name": "core-essentials",
+      "description": "Core cross-tool utilities, including delegation to the Antigravity CLI (agy) for grounded research and large multi-file audits.",
+      "author": {
+        "name": "Rootiest"
+      },
       "source": "./dist/claude-code/core-essentials"
     },
     {
       "name": "date-time",
+      "description": "Retrieves the exact current date and time for time-sensitive queries, scheduling, duration calculations, and validating chronological context.",
+      "author": {
+        "name": "Rootiest"
+      },
       "source": "./dist/claude-code/date-time"
     },
     {
       "name": "docs-sync-audit",
+      "description": "Analyzes repository delta since the last documentation update and synchronizes the Single Source of Truth (docs/, wiki, or README) with the current codebase state.",
+      "author": {
+        "name": "Rootiest"
+      },
       "source": "./dist/claude-code/docs-sync-audit"
     },
     {
       "name": "git-publish-workflow",
+      "description": "Automates branching, conventional commits, testing, and PR creation for uncommitted or staged work. Handles both independent and stacked PRs.",
+      "author": {
+        "name": "Rootiest"
+      },
       "source": "./dist/claude-code/git-publish-workflow"
     },
     {
       "name": "gitea-mcp",
+      "description": "Connects Claude Code/agy to a Gitea instance via the official gitea-mcp server, authenticated with GITEA_HOST/GITEA_ACCESS_TOKEN you set locally.",
+      "author": {
+        "name": "Rootiest"
+      },
       "source": "./dist/claude-code/gitea-mcp"
     },
     {
       "name": "github-mcp",
+      "description": "Connects Claude Code/agy to GitHub via the official MCP server, authenticated with a GITHUB_PERSONAL_ACCESS_TOKEN you set locally.",
+      "author": {
+        "name": "Rootiest"
+      },
       "source": "./dist/claude-code/github-mcp"
     },
     {
       "name": "mattpocock-skills",
+      "description": "Matt Pocock's agent skills for real engineering from mattpocock/skills \u2014 grilling, spec/ticket flows, TDD, code review, domain modelling and more. Installed straight from the upstream repo, not vendored here.",
+      "author": {
+        "name": "Rootiest"
+      },
       "source": {
         "source": "github",
         "repo": "mattpocock/skills"
@@ -46,10 +78,18 @@
     },
     {
       "name": "mirror-it",
+      "description": "Configures a Gitea repository to automatically push-mirror its contents to GitHub using the tea CLI and jq.",
+      "author": {
+        "name": "Rootiest"
+      },
       "source": "./dist/claude-code/mirror-it"
     },
     {
       "name": "ponytail",
+      "description": "Lazy senior dev mode from DietrichGebert/ponytail \u2014 forces the simplest, shortest solution that actually works (YAGNI, stdlib first, no unrequested abstractions). Installed straight from the upstream repo, not vendored here.",
+      "author": {
+        "name": "Rootiest"
+      },
       "source": {
         "source": "github",
         "repo": "DietrichGebert/ponytail"
@@ -57,22 +97,39 @@
     },
     {
       "name": "readme-sync-audit",
+      "description": "Analyzes repository delta since the last documentation update and synchronizes the README.md with the current codebase state.",
+      "author": {
+        "name": "Rootiest"
+      },
       "source": "./dist/claude-code/readme-sync-audit"
     },
     {
       "name": "ship-it",
+      "description": "Runs a comprehensive pre-flight audit, syncs the README, and publishes the changes to a new PR.",
+      "author": {
+        "name": "Rootiest"
+      },
       "source": "./dist/claude-code/ship-it"
     },
     {
       "name": "systematic-enumeration",
+      "description": "Forces element-by-element verification for finite sets to prevent counting errors.",
+      "author": {
+        "name": "Rootiest"
+      },
       "source": "./dist/claude-code/systematic-enumeration"
     },
     {
       "name": "technical-devlog-scribe",
+      "description": "Generates a highly structured, objective technical summary of a development session.",
+      "author": {
+        "name": "Rootiest"
+      },
       "source": "./dist/claude-code/technical-devlog-scribe"
     },
     {
       "name": "rootiest-ai-all",
+      "description": "Reusable AI skills for Claude Code and Antigravity CLI (all skills)",
       "source": "./dist/claude-code/rootiest-ai-all"
     }
   ]

README.md

@@ -41,90 +41,142 @@ from one source tree and published as a native marketplace for both tools.
 
 ### `systematic-enumeration`
 
-**Purpose:** Eliminate counting and membership errors caused by the model's tendency toward holistic pattern recognition.
+**Purpose:** Eliminate counting and membership errors caused by the model's
+tendency toward holistic pattern recognition.
 
 When invoked, the AI is required to work through three explicit phases:
 
-1. **Set Definition** — enumerate every member of the finite set before any analysis begins.
-2. **Atomic Element Testing** — test each item individually, producing a `[Item] → [Logic] → [Boolean]` record for every element. For character-level checks, the string is split into individual characters to bypass tokenization bias.
-3. **Reduction & Summation** — aggregate results and self-verify that the number of items tested exactly matches the set size; if there is a mismatch, Phase 2 restarts.
+1. **Set Definition** — enumerate every member of the finite set before any
+   analysis begins.
+2. **Atomic Element Testing** — test each item individually, producing a
+   `[Item] → [Logic] → [Boolean]` record for every element. For
+   character-level checks, the string is split into individual characters to
+   bypass tokenization bias.
+3. **Reduction & Summation** — aggregate results and self-verify that the number
+   of items tested exactly matches the set size; if there is a mismatch, Phase 2
+   restarts.
 
-**Use when:** counting characters in a string, verifying a property across a list of files or variables, or any membership test where a false negative carries real cost.
+**Use when:** counting characters in a string, verifying a property across a
+list of files or variables, or any membership test where a false negative
+carries real cost.
 
 ---
 
 ### `git-publish-workflow`
 
-**Purpose:** Automate the full lifecycle from uncommitted local work to an open Pull Request, with built-in quality gates.
+**Purpose:** Automate the full lifecycle from uncommitted local work to an open
+Pull Request, with built-in quality gates.
 
 The workflow runs three phases:
 
-1. **Scope Determination** — if staged changes exist, operate only on those; otherwise operate on all modified/untracked files.
-2. **Safe-Commit Sequence** — generate a `kebab-case` branch name, compose a [Conventional Commits](https://www.conventionalcommits.org/) message, then run the project's primary test/lint/build command. If verification fails, the sequence stops — nothing is pushed.
-3. **Remote Integration** — push the branch, open a PR against the default branch, and populate the description with a "Why/What" summary and a manual verification checklist (`- [ ]` items).
+1. **Scope Determination** — if staged changes exist, operate only on those;
+   otherwise operate on all modified/untracked files.
+2. **Safe-Commit Sequence** — generate a `kebab-case` branch name, compose a
+   [Conventional Commits](https://www.conventionalcommits.org/) message, then
+   run the project's primary test/lint/build command. If verification fails, the
+   sequence stops — nothing is pushed.
+3. **Remote Integration** — push the branch, open a PR against the default
+   branch, and populate the description with a "Why/What" summary and a manual
+   verification checklist (`- [ ]` items).
 
-**Use when:** you say "Ship this," "Make a PR," or invoke `/git-publish-workflow`.
+**Use when:** you say "Ship this," "Make a PR," or invoke
+`/git-publish-workflow`.
 
 ---
 
 ### `readme-sync-audit`
 
-**Purpose:** Keep `README.md` accurate by programmatically aligning it with the current state of the codebase.
+**Purpose:** Keep `README.md` accurate by programmatically aligning it with the
+current state of the codebase.
 
 Three-phase execution:
 
-1. **Delta Analysis** — locate the last commit that touched `README.md`, diff all code changes from that point to `HEAD`, and extract undocumented environment variables, CLI flags, or API changes.
-2. **Pruning & Update Audit** — remove stale setup steps or "Coming Soon" notes, correct version numbers and file paths, and synthesize documentation for newly discovered features.
-3. **Structural Integrity Check** — verify that Quick Start commands still work, all config keys are listed, and code snippets match the current API.
+1. **Delta Analysis** — locate the last commit that touched `README.md`, diff
+   all code changes from that point to `HEAD`, and extract undocumented
+   environment variables, CLI flags, or API changes.
+2. **Pruning & Update Audit** — remove stale setup steps or "Coming Soon" notes,
+   correct version numbers and file paths, and synthesize documentation for
+   newly discovered features.
+3. **Structural Integrity Check** — verify that Quick Start commands still work,
+   all config keys are listed, and code snippets match the current API.
 
-**Use when:** you say "Update the docs," "Sync the README," or invoke `/readme-sync-audit`.
+**Use when:** you say "Update the docs," "Sync the README," or invoke
+`/readme-sync-audit`.
 
 ---
 
 ### `docs-sync-audit`
 
-**Purpose:** Keep documentation accurate by aligning the project's Single Source of Truth (SSoT) with the current state of the codebase.
+**Purpose:** Keep documentation accurate by aligning the project's Single
+Source of Truth (SSoT) with the current state of the codebase.
 
-Unlike `readme-sync-audit`, this skill first discovers the documentation root. If a dedicated `docs/`, `wiki/`, or site config (e.g. `mkdocs.yml`) exists, it is treated as the SSoT and receives the detailed updates; the `README.md` is then kept as a high-level landing page that links into it. If no docs directory exists, it falls back to `README.md` as the SSoT.
+Unlike `readme-sync-audit`, this skill first discovers the documentation root.
+If a dedicated `docs/`, `wiki/`, or site config (e.g. `mkdocs.yml`) exists, it is
+treated as the SSoT and receives the detailed updates; the `README.md` is then
+kept as a high-level landing page that links into it. If no docs directory
+exists, it falls back to `README.md` as the SSoT.
 
 Three-phase execution:
 
-1. **Delta Analysis** — locate the last commit that touched the SSoT, diff all code changes to `HEAD`, and extract undocumented environment variables, CLI flags, or API changes.
-2. **Pruning & Update Audit** — route detailed changes to the correct files, remove stale steps, correct versions and paths, and synthesize docs for new features.
-3. **Structural Integrity Check** — verify Quick Start commands, configuration keys, and usage examples still match the implementation.
+1. **Delta Analysis** — locate the last commit that touched the SSoT, diff all
+   code changes to `HEAD`, and extract undocumented environment variables, CLI
+   flags, or API changes.
+2. **Pruning & Update Audit** — route detailed changes to the correct files,
+   remove stale steps, correct versions and paths, and synthesize docs for new
+   features.
+3. **Structural Integrity Check** — verify Quick Start commands, configuration
+   keys, and usage examples still match the implementation.
 
-**Use when:** you say "Update the docs," "Sync the wiki," or invoke `/docs-sync-audit`.
+**Use when:** you say "Update the docs," "Sync the wiki," or invoke
+`/docs-sync-audit`.
 
 ---
 
 ### `date-time`
 
-**Purpose:** Retrieve the exact, real-time current date and time when a task depends on the present moment.
+**Purpose:** Retrieve the exact, real-time current date and time when a task
+depends on the present moment.
 
-The skill runs the system `date` command rather than guessing, then uses the result for time-sensitive reasoning.
+The skill runs the system `date` command rather than guessing, then uses the
+result for time-sensitive reasoning.
 
-**Use when:** the request asks for the current date/time, uses relative expressions ("today", "next week", "recently"), or needs an age, duration, countdown, or check of whether an event has already occurred.
+**Use when:** the request asks for the current date/time, uses relative
+expressions ("today", "next week", "recently"), or needs an age, duration,
+countdown, or check of whether an event has already occurred.
 
 ---
 
 ### `technical-devlog-scribe`
 
-**Purpose:** Produce a dense, objective technical summary of a development session as a durable historical record optimized for future context loading.
+**Purpose:** Produce a dense, objective technical summary of a development
+session as a durable historical record optimized for future context loading.
 
-The skill writes a structured Markdown file to `AGENTS/devlogs/<kebab-case-short-description>.md`, with YAML frontmatter (date, title, tags, status) followed by the session summary. It relies strictly on what happened in the session — no hallucinated external constraints — and avoids conversational filler.
+The skill writes a structured Markdown file to
+`AGENTS/devlogs/<kebab-case-short-description>.md`, with YAML frontmatter (date,
+title, tags, status) followed by the session summary. It relies strictly on
+what happened in the session — no hallucinated external constraints — and
+avoids conversational filler.
 
-**Use when:** you say "session wrap-up," "write a devlog," "done for the day," or want an auditable record of what changed and why.
+**Use when:** you say "session wrap-up," "write a devlog," "done for the day,"
+or want an auditable record of what changed and why.
 
 ---
 
 ### `ship-it`
 
-**Purpose:** Run a comprehensive pre-flight audit and publish changes to a new PR in a single command.
+**Purpose:** Run a comprehensive pre-flight audit and publish changes to a new
+PR in a single command.
 
 Two sequential phases — Phase 2 is blocked until Phase 1 succeeds:
 
-1. **Documentation Sync & Code Audit** — acts as `/docs-sync-audit`: scans all file changes since the last documentation edit, updates the docs/README to match the current codebase, and audits code files for syntax errors or regressions. **Halts the entire workflow** if any breaking issue is found.
-2. **Git Publish Workflow** — acts as `/git-publish-workflow`: creates a descriptively named branch, commits all pending changes (including the README updates from Phase 1), pushes to the remote, and opens a Pull Request against `main`.
+1. **Documentation Sync & Code Audit** — acts as `/docs-sync-audit`: scans all
+   file changes since the last documentation edit, updates the docs/README to
+   match the current codebase, and audits code files for syntax errors or
+   regressions. **Halts the entire workflow** if any breaking issue is found.
+2. **Git Publish Workflow** — acts as `/git-publish-workflow`: creates a
+   descriptively named branch, commits all pending changes (including the
+   README updates from Phase 1), pushes to the remote, and opens a Pull Request
+   against `main`.
 
 **Use when:** you say "Ship this," "Publish my changes," or invoke `/ship-it`.
 
@@ -132,11 +184,19 @@ Two sequential phases — Phase 2 is blocked until Phase 1 succeeds:
 
 ### `mirror-it`
 
-**Purpose:** Register a Gitea push mirror so a repository's commits are automatically forwarded to a corresponding GitHub repository.
+**Purpose:** Register a Gitea push mirror so a repository's commits are
+automatically forwarded to a corresponding GitHub repository.
 
-Uses `tea api` against Gitea's `push_mirrors` endpoint, with credentials passed as `remote_username`/`remote_password` fields — never embedded in the mirror URL. Defaults to an `8h` sync interval with sync-on-commit enabled; both are overridable per request, as is the destination repository name. `GITHUB_USER`/`GITHUB_TOKEN` can come from the environment or, as a fallback, a repo-root `.env` file. All shell snippets are POSIX-compatible so the skill works under bash, zsh, or any shell the agent's tool happens to invoke — not just Fish.
+Uses `tea api` against Gitea's `push_mirrors` endpoint, with credentials passed
+as `remote_username`/`remote_password` fields — never embedded in the mirror
+URL. Defaults to an `8h` sync interval with sync-on-commit enabled; both are
+overridable per request, as is the destination repository name.
+`GITHUB_USER`/`GITHUB_TOKEN` can come from the environment or, as a fallback, a
+repo-root `.env` file.
 
-**Use when:** you say "Mirror this repository to GitHub," or invoke `/mirror-it`. Only relevant for repositories hosted on Gitea — GitHub can't push-mirror to itself.
+**Use when:** you say "Mirror this repository to GitHub," or invoke
+`/mirror-it`. Only relevant for repositories hosted on Gitea — GitHub can't
+push-mirror to itself.
 
 ---
 

dist/agy/mirror-it/skills/mirror-it/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: mirror-it
 description: Configures a Gitea repository to automatically push-mirror its contents to GitHub using the `tea` CLI and `jq`.
-version: 1.0.0
+version: 1.1.0
 user-invocable: true
 author: Rootiest
 ---
@@ -40,6 +40,10 @@ Activate this skill ONLY when one of the following is met:
 - `$GITHUB_USER` and `$GITHUB_TOKEN` (a GitHub PAT with `repo` scope) are available before step 3
   below — either already exported in the environment, or defined in a `.env` file at the
   repository root (see step 1).
+- Optionally, one of the following for the target-repo check/creation in step 4: the `gh` CLI
+  (authenticated — `gh auth status`), a GitHub MCP server, or plain `$GITHUB_TOKEN` used directly
+  against the GitHub REST API via `curl`. The first of these that's available is used; if none
+  are, step 4 degrades to a warning instead of blocking the skill.
 
 ## Defaults & Overrides
 Unless the user says otherwise, assume:
@@ -64,9 +68,48 @@ name 'exmpl_cfg'" → `$TARGET_REPO` becomes `exmpl_cfg`, `$SYNC_ON_COMMIT` beco
 2. Identify the current Gitea repository's owner (`$OWNER`) and name (`$SOURCE_REPO`).
 3. Determine `$TARGET_REPO`, `$INTERVAL`, and `$SYNC_ON_COMMIT` from the defaults above,
    applying any user overrides.
-4. Run the following snippet, substituting those values. Credentials are passed as separate
-   `remote_username`/`remote_password` fields — never embedded in the mirror URL — per Gitea's
-   `CreatePushMirrorOption` schema.
+4. **Ensure the GitHub target repository exists.** GitHub, unlike Gitea, does not create a
+   repository on first push — if `$GITHUB_USER/$TARGET_REPO` doesn't exist yet, the mirror will
+   register successfully but every sync attempt will fail. Check for the repo and create it if
+   missing, using whichever of these is available (in order of preference):
+
+   - **`gh` CLI** (preferred if authenticated — `gh auth status`):
+     ```sh
+     if ! gh repo view "$GITHUB_USER/$TARGET_REPO" >/dev/null 2>&1; then
+       gh repo create "$GITHUB_USER/$TARGET_REPO" --private --description "Push mirror of $OWNER/$SOURCE_REPO"
+     fi
+     ```
+     Match visibility to the source repo where you can determine it (e.g. via `tea api
+     "repos/$OWNER/$SOURCE_REPO"` → `.private`); otherwise default to `--private` and mention the
+     choice when reporting the outcome.
+   - **GitHub MCP server**, if one is connected: use its repo-lookup tool to check for
+     `$GITHUB_USER/$TARGET_REPO`, and its repo-creation tool to create it (private by default) if
+     absent.
+   - **Plain REST API via `curl`**, using `$GITHUB_TOKEN` directly, if neither of the above is
+     available:
+     ```sh
+     status=$(curl -s -o /dev/null -w '%{http_code}' \
+       -H "Authorization: Bearer $GITHUB_TOKEN" \
+       "https://api.github.com/repos/$GITHUB_USER/$TARGET_REPO")
+     if [ "$status" = "404" ]; then
+       curl -s -X POST \
+         -H "Authorization: Bearer $GITHUB_TOKEN" \
+         -H "Accept: application/vnd.github+json" \
+         https://api.github.com/user/repos \
+         -d "$(jq -n --arg name "$TARGET_REPO" '{name: $name, private: true}')" \
+         >/dev/null
+     fi
+     ```
+   - **If none of these are available**, don't block on it — proceed to step 5, but warn the user
+     clearly that they must create `$GITHUB_USER/$TARGET_REPO` on GitHub themselves before the
+     mirror's first sync will succeed.
+
+   If a repo already exists at the target address, leave it as-is and continue — do not treat this
+   as an error.
+
+5. Run the following snippet, substituting the values from steps 2–3. Credentials are passed as
+   separate `remote_username`/`remote_password` fields — never embedded in the mirror URL — per
+   Gitea's `CreatePushMirrorOption` schema.
 
 ```sh
 # $OWNER / $SOURCE_REPO identify the local Gitea API endpoint.
@@ -85,4 +128,6 @@ payload=$(jq -n \
 tea api --method POST --data "$payload" "repos/$OWNER/$SOURCE_REPO/push_mirrors"
 ```
 
-5. Report the outcome to the user, without echoing the request body, `.env` contents, or token.
+6. Report the outcome to the user, without echoing the request body, `.env` contents, or token —
+   including whether the GitHub repo already existed or was just created, and, if step 4 had no
+   capability to check/create it, a reminder that they need to create it manually.

dist/agy/rootiest-ai-all/skills/mirror-it/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: mirror-it
 description: Configures a Gitea repository to automatically push-mirror its contents to GitHub using the `tea` CLI and `jq`.
-version: 1.0.0
+version: 1.1.0
 user-invocable: true
 author: Rootiest
 ---
@@ -40,6 +40,10 @@ Activate this skill ONLY when one of the following is met:
 - `$GITHUB_USER` and `$GITHUB_TOKEN` (a GitHub PAT with `repo` scope) are available before step 3
   below — either already exported in the environment, or defined in a `.env` file at the
   repository root (see step 1).
+- Optionally, one of the following for the target-repo check/creation in step 4: the `gh` CLI
+  (authenticated — `gh auth status`), a GitHub MCP server, or plain `$GITHUB_TOKEN` used directly
+  against the GitHub REST API via `curl`. The first of these that's available is used; if none
+  are, step 4 degrades to a warning instead of blocking the skill.
 
 ## Defaults & Overrides
 Unless the user says otherwise, assume:
@@ -64,9 +68,48 @@ name 'exmpl_cfg'" → `$TARGET_REPO` becomes `exmpl_cfg`, `$SYNC_ON_COMMIT` beco
 2. Identify the current Gitea repository's owner (`$OWNER`) and name (`$SOURCE_REPO`).
 3. Determine `$TARGET_REPO`, `$INTERVAL`, and `$SYNC_ON_COMMIT` from the defaults above,
    applying any user overrides.
-4. Run the following snippet, substituting those values. Credentials are passed as separate
-   `remote_username`/`remote_password` fields — never embedded in the mirror URL — per Gitea's
-   `CreatePushMirrorOption` schema.
+4. **Ensure the GitHub target repository exists.** GitHub, unlike Gitea, does not create a
+   repository on first push — if `$GITHUB_USER/$TARGET_REPO` doesn't exist yet, the mirror will
+   register successfully but every sync attempt will fail. Check for the repo and create it if
+   missing, using whichever of these is available (in order of preference):
+
+   - **`gh` CLI** (preferred if authenticated — `gh auth status`):
+     ```sh
+     if ! gh repo view "$GITHUB_USER/$TARGET_REPO" >/dev/null 2>&1; then
+       gh repo create "$GITHUB_USER/$TARGET_REPO" --private --description "Push mirror of $OWNER/$SOURCE_REPO"
+     fi
+     ```
+     Match visibility to the source repo where you can determine it (e.g. via `tea api
+     "repos/$OWNER/$SOURCE_REPO"` → `.private`); otherwise default to `--private` and mention the
+     choice when reporting the outcome.
+   - **GitHub MCP server**, if one is connected: use its repo-lookup tool to check for
+     `$GITHUB_USER/$TARGET_REPO`, and its repo-creation tool to create it (private by default) if
+     absent.
+   - **Plain REST API via `curl`**, using `$GITHUB_TOKEN` directly, if neither of the above is
+     available:
+     ```sh
+     status=$(curl -s -o /dev/null -w '%{http_code}' \
+       -H "Authorization: Bearer $GITHUB_TOKEN" \
+       "https://api.github.com/repos/$GITHUB_USER/$TARGET_REPO")
+     if [ "$status" = "404" ]; then
+       curl -s -X POST \
+         -H "Authorization: Bearer $GITHUB_TOKEN" \
+         -H "Accept: application/vnd.github+json" \
+         https://api.github.com/user/repos \
+         -d "$(jq -n --arg name "$TARGET_REPO" '{name: $name, private: true}')" \
+         >/dev/null
+     fi
+     ```
+   - **If none of these are available**, don't block on it — proceed to step 5, but warn the user
+     clearly that they must create `$GITHUB_USER/$TARGET_REPO` on GitHub themselves before the
+     mirror's first sync will succeed.
+
+   If a repo already exists at the target address, leave it as-is and continue — do not treat this
+   as an error.
+
+5. Run the following snippet, substituting the values from steps 2–3. Credentials are passed as
+   separate `remote_username`/`remote_password` fields — never embedded in the mirror URL — per
+   Gitea's `CreatePushMirrorOption` schema.
 
 ```sh
 # $OWNER / $SOURCE_REPO identify the local Gitea API endpoint.
@@ -85,4 +128,6 @@ payload=$(jq -n \
 tea api --method POST --data "$payload" "repos/$OWNER/$SOURCE_REPO/push_mirrors"
 ```
 
-5. Report the outcome to the user, without echoing the request body, `.env` contents, or token.
+6. Report the outcome to the user, without echoing the request body, `.env` contents, or token —
+   including whether the GitHub repo already existed or was just created, and, if step 4 had no
+   capability to check/create it, a reminder that they need to create it manually.

dist/claude-code/core-essentials/.claude-plugin/plugin.json

@@ -2,5 +2,7 @@
   "name": "core-essentials",
   "description": "Core cross-tool utilities, including delegation to the Antigravity CLI (agy) for grounded research and large multi-file audits.",
   "version": "1.0.0",
-  "author": "Rootiest"
+  "author": {
+    "name": "Rootiest"
+  }
 }

dist/claude-code/date-time/.claude-plugin/plugin.json

@@ -2,5 +2,7 @@
   "name": "date-time",
   "description": "Retrieves the exact current date and time for time-sensitive queries, scheduling, duration calculations, and validating chronological context.",
   "version": "1.0.0",
-  "author": "Rootiest"
+  "author": {
+    "name": "Rootiest"
+  }
 }

dist/claude-code/docs-sync-audit/.claude-plugin/plugin.json

@@ -2,5 +2,7 @@
   "name": "docs-sync-audit",
   "description": "Analyzes repository delta since the last documentation update and synchronizes the Single Source of Truth (docs/, wiki, or README) with the current codebase state.",
   "version": "1.3.0",
-  "author": "Rootiest"
+  "author": {
+    "name": "Rootiest"
+  }
 }

dist/claude-code/git-publish-workflow/.claude-plugin/plugin.json

@@ -2,5 +2,7 @@
   "name": "git-publish-workflow",
   "description": "Automates branching, conventional commits, testing, and PR creation for uncommitted or staged work. Handles both independent and stacked PRs.",
   "version": "1.2.0",
-  "author": "Rootiest"
+  "author": {
+    "name": "Rootiest"
+  }
 }

dist/claude-code/gitea-mcp/.claude-plugin/plugin.json

@@ -2,5 +2,7 @@
   "name": "gitea-mcp",
   "description": "Connects Claude Code/agy to a Gitea instance via the official gitea-mcp server, authenticated with GITEA_HOST/GITEA_ACCESS_TOKEN you set locally.",
   "version": "1.0.0",
-  "author": "Rootiest"
+  "author": {
+    "name": "Rootiest"
+  }
 }

dist/claude-code/github-mcp/.claude-plugin/plugin.json

@@ -2,5 +2,7 @@
   "name": "github-mcp",
   "description": "Connects Claude Code/agy to GitHub via the official MCP server, authenticated with a GITHUB_PERSONAL_ACCESS_TOKEN you set locally.",
   "version": "1.0.0",
-  "author": "Rootiest"
+  "author": {
+    "name": "Rootiest"
+  }
 }

dist/claude-code/mirror-it/.claude-plugin/plugin.json

@@ -1,6 +1,8 @@
 {
   "name": "mirror-it",
   "description": "Configures a Gitea repository to automatically push-mirror its contents to GitHub using the tea CLI and jq.",
-  "version": "1.0.0",
-  "author": "Rootiest"
+  "version": "1.1.0",
+  "author": {
+    "name": "Rootiest"
+  }
 }

dist/claude-code/mirror-it/skills/mirror-it/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: mirror-it
 description: Configures a Gitea repository to automatically push-mirror its contents to GitHub using the `tea` CLI and `jq`.
-version: 1.0.0
+version: 1.1.0
 user-invocable: true
 author: Rootiest
 ---
@@ -40,6 +40,10 @@ Activate this skill ONLY when one of the following is met:
 - `$GITHUB_USER` and `$GITHUB_TOKEN` (a GitHub PAT with `repo` scope) are available before step 3
   below — either already exported in the environment, or defined in a `.env` file at the
   repository root (see step 1).
+- Optionally, one of the following for the target-repo check/creation in step 4: the `gh` CLI
+  (authenticated — `gh auth status`), a GitHub MCP server, or plain `$GITHUB_TOKEN` used directly
+  against the GitHub REST API via `curl`. The first of these that's available is used; if none
+  are, step 4 degrades to a warning instead of blocking the skill.
 
 ## Defaults & Overrides
 Unless the user says otherwise, assume:
@@ -64,9 +68,48 @@ name 'exmpl_cfg'" → `$TARGET_REPO` becomes `exmpl_cfg`, `$SYNC_ON_COMMIT` beco
 2. Identify the current Gitea repository's owner (`$OWNER`) and name (`$SOURCE_REPO`).
 3. Determine `$TARGET_REPO`, `$INTERVAL`, and `$SYNC_ON_COMMIT` from the defaults above,
    applying any user overrides.
-4. Run the following snippet, substituting those values. Credentials are passed as separate
-   `remote_username`/`remote_password` fields — never embedded in the mirror URL — per Gitea's
-   `CreatePushMirrorOption` schema.
+4. **Ensure the GitHub target repository exists.** GitHub, unlike Gitea, does not create a
+   repository on first push — if `$GITHUB_USER/$TARGET_REPO` doesn't exist yet, the mirror will
+   register successfully but every sync attempt will fail. Check for the repo and create it if
+   missing, using whichever of these is available (in order of preference):
+
+   - **`gh` CLI** (preferred if authenticated — `gh auth status`):
+     ```sh
+     if ! gh repo view "$GITHUB_USER/$TARGET_REPO" >/dev/null 2>&1; then
+       gh repo create "$GITHUB_USER/$TARGET_REPO" --private --description "Push mirror of $OWNER/$SOURCE_REPO"
+     fi
+     ```
+     Match visibility to the source repo where you can determine it (e.g. via `tea api
+     "repos/$OWNER/$SOURCE_REPO"` → `.private`); otherwise default to `--private` and mention the
+     choice when reporting the outcome.
+   - **GitHub MCP server**, if one is connected: use its repo-lookup tool to check for
+     `$GITHUB_USER/$TARGET_REPO`, and its repo-creation tool to create it (private by default) if
+     absent.
+   - **Plain REST API via `curl`**, using `$GITHUB_TOKEN` directly, if neither of the above is
+     available:
+     ```sh
+     status=$(curl -s -o /dev/null -w '%{http_code}' \
+       -H "Authorization: Bearer $GITHUB_TOKEN" \
+       "https://api.github.com/repos/$GITHUB_USER/$TARGET_REPO")
+     if [ "$status" = "404" ]; then
+       curl -s -X POST \
+         -H "Authorization: Bearer $GITHUB_TOKEN" \
+         -H "Accept: application/vnd.github+json" \
+         https://api.github.com/user/repos \
+         -d "$(jq -n --arg name "$TARGET_REPO" '{name: $name, private: true}')" \
+         >/dev/null
+     fi
+     ```
+   - **If none of these are available**, don't block on it — proceed to step 5, but warn the user
+     clearly that they must create `$GITHUB_USER/$TARGET_REPO` on GitHub themselves before the
+     mirror's first sync will succeed.
+
+   If a repo already exists at the target address, leave it as-is and continue — do not treat this
+   as an error.
+
+5. Run the following snippet, substituting the values from steps 2–3. Credentials are passed as
+   separate `remote_username`/`remote_password` fields — never embedded in the mirror URL — per
+   Gitea's `CreatePushMirrorOption` schema.
 
 ```sh
 # $OWNER / $SOURCE_REPO identify the local Gitea API endpoint.
@@ -85,4 +128,6 @@ payload=$(jq -n \
 tea api --method POST --data "$payload" "repos/$OWNER/$SOURCE_REPO/push_mirrors"
 ```
 
-5. Report the outcome to the user, without echoing the request body, `.env` contents, or token.
+6. Report the outcome to the user, without echoing the request body, `.env` contents, or token —
+   including whether the GitHub repo already existed or was just created, and, if step 4 had no
+   capability to check/create it, a reminder that they need to create it manually.

dist/claude-code/readme-sync-audit/.claude-plugin/plugin.json

@@ -2,5 +2,7 @@
   "name": "readme-sync-audit",
   "description": "Analyzes repository delta since the last documentation update and synchronizes the README.md with the current codebase state.",
   "version": "1.2.0",
-  "author": "Rootiest"
+  "author": {
+    "name": "Rootiest"
+  }
 }

dist/claude-code/rootiest-ai-all/skills/mirror-it/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: mirror-it
 description: Configures a Gitea repository to automatically push-mirror its contents to GitHub using the `tea` CLI and `jq`.
-version: 1.0.0
+version: 1.1.0
 user-invocable: true
 author: Rootiest
 ---
@@ -40,6 +40,10 @@ Activate this skill ONLY when one of the following is met:
 - `$GITHUB_USER` and `$GITHUB_TOKEN` (a GitHub PAT with `repo` scope) are available before step 3
   below — either already exported in the environment, or defined in a `.env` file at the
   repository root (see step 1).
+- Optionally, one of the following for the target-repo check/creation in step 4: the `gh` CLI
+  (authenticated — `gh auth status`), a GitHub MCP server, or plain `$GITHUB_TOKEN` used directly
+  against the GitHub REST API via `curl`. The first of these that's available is used; if none
+  are, step 4 degrades to a warning instead of blocking the skill.
 
 ## Defaults & Overrides
 Unless the user says otherwise, assume:
@@ -64,9 +68,48 @@ name 'exmpl_cfg'" → `$TARGET_REPO` becomes `exmpl_cfg`, `$SYNC_ON_COMMIT` beco
 2. Identify the current Gitea repository's owner (`$OWNER`) and name (`$SOURCE_REPO`).
 3. Determine `$TARGET_REPO`, `$INTERVAL`, and `$SYNC_ON_COMMIT` from the defaults above,
    applying any user overrides.
-4. Run the following snippet, substituting those values. Credentials are passed as separate
-   `remote_username`/`remote_password` fields — never embedded in the mirror URL — per Gitea's
-   `CreatePushMirrorOption` schema.
+4. **Ensure the GitHub target repository exists.** GitHub, unlike Gitea, does not create a
+   repository on first push — if `$GITHUB_USER/$TARGET_REPO` doesn't exist yet, the mirror will
+   register successfully but every sync attempt will fail. Check for the repo and create it if
+   missing, using whichever of these is available (in order of preference):
+
+   - **`gh` CLI** (preferred if authenticated — `gh auth status`):
+     ```sh
+     if ! gh repo view "$GITHUB_USER/$TARGET_REPO" >/dev/null 2>&1; then
+       gh repo create "$GITHUB_USER/$TARGET_REPO" --private --description "Push mirror of $OWNER/$SOURCE_REPO"
+     fi
+     ```
+     Match visibility to the source repo where you can determine it (e.g. via `tea api
+     "repos/$OWNER/$SOURCE_REPO"` → `.private`); otherwise default to `--private` and mention the
+     choice when reporting the outcome.
+   - **GitHub MCP server**, if one is connected: use its repo-lookup tool to check for
+     `$GITHUB_USER/$TARGET_REPO`, and its repo-creation tool to create it (private by default) if
+     absent.
+   - **Plain REST API via `curl`**, using `$GITHUB_TOKEN` directly, if neither of the above is
+     available:
+     ```sh
+     status=$(curl -s -o /dev/null -w '%{http_code}' \
+       -H "Authorization: Bearer $GITHUB_TOKEN" \
+       "https://api.github.com/repos/$GITHUB_USER/$TARGET_REPO")
+     if [ "$status" = "404" ]; then
+       curl -s -X POST \
+         -H "Authorization: Bearer $GITHUB_TOKEN" \
+         -H "Accept: application/vnd.github+json" \
+         https://api.github.com/user/repos \
+         -d "$(jq -n --arg name "$TARGET_REPO" '{name: $name, private: true}')" \
+         >/dev/null
+     fi
+     ```
+   - **If none of these are available**, don't block on it — proceed to step 5, but warn the user
+     clearly that they must create `$GITHUB_USER/$TARGET_REPO` on GitHub themselves before the
+     mirror's first sync will succeed.
+
+   If a repo already exists at the target address, leave it as-is and continue — do not treat this
+   as an error.
+
+5. Run the following snippet, substituting the values from steps 2–3. Credentials are passed as
+   separate `remote_username`/`remote_password` fields — never embedded in the mirror URL — per
+   Gitea's `CreatePushMirrorOption` schema.
 
 ```sh
 # $OWNER / $SOURCE_REPO identify the local Gitea API endpoint.
@@ -85,4 +128,6 @@ payload=$(jq -n \
 tea api --method POST --data "$payload" "repos/$OWNER/$SOURCE_REPO/push_mirrors"
 ```
 
-5. Report the outcome to the user, without echoing the request body, `.env` contents, or token.
+6. Report the outcome to the user, without echoing the request body, `.env` contents, or token —
+   including whether the GitHub repo already existed or was just created, and, if step 4 had no
+   capability to check/create it, a reminder that they need to create it manually.

dist/claude-code/ship-it/.claude-plugin/plugin.json

@@ -2,5 +2,7 @@
   "name": "ship-it",
   "description": "Runs a comprehensive pre-flight audit, syncs the README, and publishes the changes to a new PR.",
   "version": "1.0.0",
-  "author": "Rootiest"
+  "author": {
+    "name": "Rootiest"
+  }
 }

dist/claude-code/systematic-enumeration/.claude-plugin/plugin.json

@@ -2,5 +2,7 @@
   "name": "systematic-enumeration",
   "description": "Forces element-by-element verification for finite sets to prevent counting errors.",
   "version": "1.0.0",
-  "author": "Rootiest"
+  "author": {
+    "name": "Rootiest"
+  }
 }

dist/claude-code/technical-devlog-scribe/.claude-plugin/plugin.json

@@ -2,5 +2,7 @@
   "name": "technical-devlog-scribe",
   "description": "Generates a highly structured, objective technical summary of a development session.",
   "version": "1.0.0",
-  "author": "Rootiest"
+  "author": {
+    "name": "Rootiest"
+  }
 }

plugins/caveman/plugin.json

@@ -2,7 +2,7 @@
   "name": "caveman",
   "description": "Ultra-compressed communication mode from JuliusBrussee/caveman — cuts output tokens while keeping full technical accuracy. Installed straight from the upstream repo, not vendored here.",
   "version": "1.0.0",
-  "author": "Rootiest",
+  "author": { "name": "Rootiest" },
   "targets": ["claude-code"],
   "external": {
     "claude-code": {

plugins/core-essentials/plugin.json

@@ -2,5 +2,5 @@
   "name": "core-essentials",
   "description": "Core cross-tool utilities, including delegation to the Antigravity CLI (agy) for grounded research and large multi-file audits.",
   "version": "1.0.0",
-  "author": "Rootiest"
+  "author": { "name": "Rootiest" }
 }

plugins/date-time/plugin.json

@@ -2,5 +2,5 @@
   "name": "date-time",
   "description": "Retrieves the exact current date and time for time-sensitive queries, scheduling, duration calculations, and validating chronological context.",
   "version": "1.0.0",
-  "author": "Rootiest"
+  "author": { "name": "Rootiest" }
 }

plugins/docs-sync-audit/plugin.json

@@ -2,5 +2,5 @@
   "name": "docs-sync-audit",
   "description": "Analyzes repository delta since the last documentation update and synchronizes the Single Source of Truth (docs/, wiki, or README) with the current codebase state.",
   "version": "1.3.0",
-  "author": "Rootiest"
+  "author": { "name": "Rootiest" }
 }

plugins/git-publish-workflow/plugin.json

@@ -2,5 +2,5 @@
   "name": "git-publish-workflow",
   "description": "Automates branching, conventional commits, testing, and PR creation for uncommitted or staged work. Handles both independent and stacked PRs.",
   "version": "1.2.0",
-  "author": "Rootiest"
+  "author": { "name": "Rootiest" }
 }

plugins/gitea-mcp/plugin.json

@@ -2,5 +2,5 @@
   "name": "gitea-mcp",
   "description": "Connects Claude Code/agy to a Gitea instance via the official gitea-mcp server, authenticated with GITEA_HOST/GITEA_ACCESS_TOKEN you set locally.",
   "version": "1.0.0",
-  "author": "Rootiest"
+  "author": { "name": "Rootiest" }
 }

plugins/github-mcp/plugin.json

@@ -2,5 +2,5 @@
   "name": "github-mcp",
   "description": "Connects Claude Code/agy to GitHub via the official MCP server, authenticated with a GITHUB_PERSONAL_ACCESS_TOKEN you set locally.",
   "version": "1.0.0",
-  "author": "Rootiest"
+  "author": { "name": "Rootiest" }
 }

plugins/mattpocock-skills/plugin.json

@@ -2,7 +2,7 @@
   "name": "mattpocock-skills",
   "description": "Matt Pocock's agent skills for real engineering from mattpocock/skills — grilling, spec/ticket flows, TDD, code review, domain modelling and more. Installed straight from the upstream repo, not vendored here.",
   "version": "1.0.0",
-  "author": "Rootiest",
+  "author": { "name": "Rootiest" },
   "targets": ["claude-code"],
   "external": {
     "claude-code": {

plugins/mirror-it/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "mirror-it",
   "description": "Configures a Gitea repository to automatically push-mirror its contents to GitHub using the tea CLI and jq.",
-  "version": "1.0.0",
-  "author": "Rootiest"
+  "version": "1.1.0",
+  "author": { "name": "Rootiest" }
 }

plugins/mirror-it/skills/mirror-it/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: mirror-it
 description: Configures a Gitea repository to automatically push-mirror its contents to GitHub using the `tea` CLI and `jq`.
-version: 1.0.0
+version: 1.1.0
 user-invocable: true
 author: Rootiest
 ---
@@ -40,6 +40,10 @@ Activate this skill ONLY when one of the following is met:
 - `$GITHUB_USER` and `$GITHUB_TOKEN` (a GitHub PAT with `repo` scope) are available before step 3
   below — either already exported in the environment, or defined in a `.env` file at the
   repository root (see step 1).
+- Optionally, one of the following for the target-repo check/creation in step 4: the `gh` CLI
+  (authenticated — `gh auth status`), a GitHub MCP server, or plain `$GITHUB_TOKEN` used directly
+  against the GitHub REST API via `curl`. The first of these that's available is used; if none
+  are, step 4 degrades to a warning instead of blocking the skill.
 
 ## Defaults & Overrides
 Unless the user says otherwise, assume:
@@ -64,9 +68,48 @@ name 'exmpl_cfg'" → `$TARGET_REPO` becomes `exmpl_cfg`, `$SYNC_ON_COMMIT` beco
 2. Identify the current Gitea repository's owner (`$OWNER`) and name (`$SOURCE_REPO`).
 3. Determine `$TARGET_REPO`, `$INTERVAL`, and `$SYNC_ON_COMMIT` from the defaults above,
    applying any user overrides.
-4. Run the following snippet, substituting those values. Credentials are passed as separate
-   `remote_username`/`remote_password` fields — never embedded in the mirror URL — per Gitea's
-   `CreatePushMirrorOption` schema.
+4. **Ensure the GitHub target repository exists.** GitHub, unlike Gitea, does not create a
+   repository on first push — if `$GITHUB_USER/$TARGET_REPO` doesn't exist yet, the mirror will
+   register successfully but every sync attempt will fail. Check for the repo and create it if
+   missing, using whichever of these is available (in order of preference):
+
+   - **`gh` CLI** (preferred if authenticated — `gh auth status`):
+     ```sh
+     if ! gh repo view "$GITHUB_USER/$TARGET_REPO" >/dev/null 2>&1; then
+       gh repo create "$GITHUB_USER/$TARGET_REPO" --private --description "Push mirror of $OWNER/$SOURCE_REPO"
+     fi
+     ```
+     Match visibility to the source repo where you can determine it (e.g. via `tea api
+     "repos/$OWNER/$SOURCE_REPO"` → `.private`); otherwise default to `--private` and mention the
+     choice when reporting the outcome.
+   - **GitHub MCP server**, if one is connected: use its repo-lookup tool to check for
+     `$GITHUB_USER/$TARGET_REPO`, and its repo-creation tool to create it (private by default) if
+     absent.
+   - **Plain REST API via `curl`**, using `$GITHUB_TOKEN` directly, if neither of the above is
+     available:
+     ```sh
+     status=$(curl -s -o /dev/null -w '%{http_code}' \
+       -H "Authorization: Bearer $GITHUB_TOKEN" \
+       "https://api.github.com/repos/$GITHUB_USER/$TARGET_REPO")
+     if [ "$status" = "404" ]; then
+       curl -s -X POST \
+         -H "Authorization: Bearer $GITHUB_TOKEN" \
+         -H "Accept: application/vnd.github+json" \
+         https://api.github.com/user/repos \
+         -d "$(jq -n --arg name "$TARGET_REPO" '{name: $name, private: true}')" \
+         >/dev/null
+     fi
+     ```
+   - **If none of these are available**, don't block on it — proceed to step 5, but warn the user
+     clearly that they must create `$GITHUB_USER/$TARGET_REPO` on GitHub themselves before the
+     mirror's first sync will succeed.
+
+   If a repo already exists at the target address, leave it as-is and continue — do not treat this
+   as an error.
+
+5. Run the following snippet, substituting the values from steps 2–3. Credentials are passed as
+   separate `remote_username`/`remote_password` fields — never embedded in the mirror URL — per
+   Gitea's `CreatePushMirrorOption` schema.
 
 ```sh
 # $OWNER / $SOURCE_REPO identify the local Gitea API endpoint.
@@ -85,4 +128,6 @@ payload=$(jq -n \
 tea api --method POST --data "$payload" "repos/$OWNER/$SOURCE_REPO/push_mirrors"
 ```
 
-5. Report the outcome to the user, without echoing the request body, `.env` contents, or token.
+6. Report the outcome to the user, without echoing the request body, `.env` contents, or token —
+   including whether the GitHub repo already existed or was just created, and, if step 4 had no
+   capability to check/create it, a reminder that they need to create it manually.

plugins/ponytail/plugin.json

@@ -2,7 +2,7 @@
   "name": "ponytail",
   "description": "Lazy senior dev mode from DietrichGebert/ponytail — forces the simplest, shortest solution that actually works (YAGNI, stdlib first, no unrequested abstractions). Installed straight from the upstream repo, not vendored here.",
   "version": "1.0.0",
-  "author": "Rootiest",
+  "author": { "name": "Rootiest" },
   "targets": ["claude-code"],
   "external": {
     "claude-code": {

plugins/readme-sync-audit/plugin.json

@@ -2,5 +2,5 @@
   "name": "readme-sync-audit",
   "description": "Analyzes repository delta since the last documentation update and synchronizes the README.md with the current codebase state.",
   "version": "1.2.0",
-  "author": "Rootiest"
+  "author": { "name": "Rootiest" }
 }

plugins/ship-it/plugin.json

@@ -2,5 +2,5 @@
   "name": "ship-it",
   "description": "Runs a comprehensive pre-flight audit, syncs the README, and publishes the changes to a new PR.",
   "version": "1.0.0",
-  "author": "Rootiest"
+  "author": { "name": "Rootiest" }
 }

plugins/systematic-enumeration/plugin.json

@@ -2,5 +2,5 @@
   "name": "systematic-enumeration",
   "description": "Forces element-by-element verification for finite sets to prevent counting errors.",
   "version": "1.0.0",
-  "author": "Rootiest"
+  "author": { "name": "Rootiest" }
 }

plugins/technical-devlog-scribe/plugin.json

@@ -2,5 +2,5 @@
   "name": "technical-devlog-scribe",
   "description": "Generates a highly structured, objective technical summary of a development session.",
   "version": "1.0.0",
-  "author": "Rootiest"
+  "author": { "name": "Rootiest" }
 }

scripts/generate_plugins.py

@@ -159,6 +159,17 @@ def validate_targets(value: "list | None", valid_targets: list[str], context: st
     return value
 
 
+def validate_author(value: "dict | str | None", context: str) -> None:
+    """Validate an optional 'author' field. The Claude Code marketplace/plugin
+    schemas both require an object ({"name": ...}), not a bare string."""
+    if value is None:
+        return
+    if not isinstance(value, dict) or not value.get("name"):
+        raise ValidationError(
+            f'{context}: \'author\' must be an object with a \'name\' field, e.g. {{"name": "..."}}'
+        )
+
+
 def validate_external(value: "dict | None", valid_targets: list[str], context: str) -> "dict | None":
     """Validate an optional 'external' map of target -> marketplace source object."""
     if value is None:
@@ -189,6 +200,7 @@ def discover_plugins(merged_root: Path, valid_targets: list[str]) -> list[dict]:
         for field in REQUIRED_PLUGIN_FIELDS:
             if not meta.get(field):
                 raise ValidationError(f"{manifest_path}: missing required field '{field}'")
+        validate_author(meta.get("author"), str(manifest_path))
         plugin_targets = validate_targets(meta.get("targets"), valid_targets, str(manifest_path))
         plugin_external = validate_external(meta.get("external"), valid_targets, str(manifest_path))
         if plugin_external is not None:
@@ -399,16 +411,25 @@ def gen_claude_code(manifest: dict, plugins: list[dict], out_dir: Path) -> None:
         shutil.rmtree(plugins_out)
     plugins_out.mkdir(parents=True)
 
+    def marketplace_entry(name: str, meta: dict, source) -> dict:
+        entry: dict = {"name": name}
+        if meta.get("description"):
+            entry["description"] = meta["description"]
+        if meta.get("author"):
+            entry["author"] = meta["author"]
+        entry["source"] = source
+        return entry
+
     marketplace_entries = []
     for p in plugins:
         name = p["meta"]["name"]
         external_source = (p["external"] or {}).get("claude-code") if p["external"] else None
         if external_source is not None:
             if plugin_supports(p, "claude-code"):
-                marketplace_entries.append({"name": name, "source": external_source})
+                marketplace_entries.append(marketplace_entry(name, p["meta"], external_source))
             continue
         if write_claude_plugin(p, plugins_out / name):
-            marketplace_entries.append({"name": name, "source": f"./dist/claude-code/{name}"})
+            marketplace_entries.append(marketplace_entry(name, p["meta"], f"./dist/claude-code/{name}"))
 
     bundle_id = manifest["bundle"]["id"]
     bundle_skills = [(p, s) for p in plugins for s in active_skills(p, "claude-code")]
@@ -426,7 +447,13 @@ def gen_claude_code(manifest: dict, plugins: list[dict], out_dir: Path) -> None:
         bundle_skills_dir = bundle_dest / "skills"
         for _, s in bundle_skills:
             shutil.copytree(s["dir"], bundle_skills_dir / s["frontmatter"]["name"])
-        marketplace_entries.append({"name": bundle_id, "source": f"./dist/claude-code/{bundle_id}"})
+        marketplace_entries.append(
+            marketplace_entry(
+                bundle_id,
+                {"description": f"{manifest['marketplace']['description']} (all skills)"},
+                f"./dist/claude-code/{bundle_id}",
+            )
+        )
 
     marketplace = {
         "name": manifest["marketplace"]["name"],