[joris-sevenlab]

What this changes

Fixes the five findings in #102 from a project on battlestack 1.4.0. Stacked on #101, which it builds on for ownership; merge #101 first.

Closes #102.

Why and what

1. Header actions slot (nuxt4:landing-shell): the landing layout's header has an empty #header-actions target, and its centre slot shows at every width (:ui="{ center: 'flex' }"). The nav menu keeps its old behaviour (hidden lg:flex, mobile menu unchanged), so this differs from the issue's snippet, which dropped both. A page adds an action with <Teleport to="#header-actions" defer>, client-side only. The battlestack-ui skill tells agents to do that instead of editing the layout.
   • The layout is structural, so an existing project owns it and pull never changes it. The pull that crosses landing-shell 1.4.3 lists the two lines to add as a step in REPORT.md, once, and only when the layout has a UHeader without the target.
2. Prompt limit (nuxt4:prompts): 100,000 characters instead of 10,000 (the column is text). The API now returns the real reason ("Content is longer than 100,000 characters") rather than h3's generic "Validation Error", and the page shows it through serverErrorMessage, as the user-admin pages already do.
3. AI e2e test (nuxt4:mastra, mastra-admin subtree): sets the model from NUXT_AI_GATEWAY_CHAT_MODEL, which every scaffold's .env defines, instead of a hard-coded openai/gpt-5.6-luna, so a run against a dev server leaves a model its gateway key can use. Skips when unset.
4. Mastra agents on boot (nuxt4:mastra): calls listAgents() (checked in @mastra/core 1.78; getAgents is gone), falls back to getAgents() for older Mastra, and warns instead of silently registering only default.
5. own on a path two features track: owning is per path now.
   • own/disown apply to every feature that tracks the path.
   • The writers, obsolete deletion, format reconciliation, doctor, add and remove treat a path owned under any feature as owned. Manifests from the old own need no migration.
   • pull never changes an owned file, edited or not: the overwrite repair puts it back.

Type of change

• Bug fix
• New feature or template
• Refactor with no behaviour change
• Documentation
• Build, CI or release tooling

AI assistance

• No AI agent involved
• Claude Code
• Cursor
• GitHub Copilot
• Codex
• Other, named below

Agent, service and model, if you know it: Claude Code, Opus 5.5

How much of this is agent-written?

• Whole change, reviewed by me before opening
• Substantial parts, with my own edits on top
• Small parts, or just the tests or docs
• Only the commit message or this description

Feature versions

• I changed what a feature emits, and bumped that feature's version: nuxt4:landing-shell 1.4.3, nuxt4:prompts 1.2.4, nuxt4:mastra 2.1.8, shared:ai-tool-config 1.3.1 (against pull: merge into edited files, generate migrations, and the review fixes #101's branch)
• I changed a collectDocs or collectEnv contribution, and also bumped the feature that writes the target file
• Not applicable

Changelog

• I added a line under ## Unreleased saying what changes for someone using battlestack
• Not applicable

Testing

pnpm tsc
pnpm test                                        # 106 files, all passing
BUMP_CHECK_BASE=origin/fix/pull-fixes pnpm bump:check

• New tests, each red without its fix:

  • own claims a path for both tracking features, and disown releases it.
  • Pull leaves a path owned under another feature untouched.
  • doctor reports no drift for it, and add respects it.
  • The overwrite repair puts back an owned file, edited or not.
  • The header-actions step is listed once, and not again.

• Prompts: a new e2e case saves a 20,000-character prompt, gets the reason for one over 100,000, and resets the prompt.

• Points 1 to 4 are template code. CI's scaffold job lints, type-checks and runs them in a real project.

• pnpm tsc passes

• pnpm test passes

• I removed the behaviour each new assertion covers, watched the test fail, and put it back

Checklist

• Scoped to one concern. Unrelated cleanups are in their own commit or PR
• Public-facing docs updated where behaviour changed (docs/, README.md)
• No internal tracker numbers, post-mortem narrative, or internal infrastructure details in the diff
• Files under packages/*/templates/** are LF
• No secrets, tokens or real credentials, including in test fixtures

Anything else

• Disowning a structural file still gets it owned again on the next pull. That predates this PR and is unchanged.
• While doing point 1 I found that bump:check missed template changes in features that pass this.id (landing-shell, build-info). That fix went into pull: merge into edited files, generate migrations, and the review fixes #101, since pull: merge into edited files, never overwrite local work #94's layout change had slipped through without a bump.