Invocation for
vc-ownership(launcherownership)Same three-path shape as the fleet, with this skill's literals — see the canonical Delegation Matrix:
Path Literal for this skill 1. User-launched worker vibecrafted ownership <agent>2. Interactive /vc-ownership— execute in this session; use native subagents when required; do not externalize merely because a launcher exists3. Agent-operator may dispatch the worker form above via vc-dispatch/ operator lines while preserving this skill's identity
Freer native on some runs ≠ abandon external fleet.
vc-dispatchandvc-shipkeep their own identities.
vc-ownership
Autonomous delivery posture. Take responsibility end-to-end, drive to green, then prove the product surface is not still undone.
Taxonomy
vc-ownership:
kind: autonomous_posture
scope: interactive_or_headless_session
meaning: take responsibility end-to-end, minimize questions, drive to green
autonomy: full
Skill invocation is not runtime invocation. If the operator says
$vc-ownership inside the current conversation, the current agent adopts the
autonomous delivery posture. A separate runtime run exists only when the
operator or framework launches vibecrafted ownership <agent> ....
See TAXONOMY.md for the posture/runtime split.
Living Tree / Worktree Rule
This workflow runs in the operator's current checkout and current branch. Do not create, switch to, or move execution into a git worktree unless the operator explicitly asks for a worktree in this prompt. Generic words like "isolate", " parallel", or "clean branch" are not enough. Re-read files before editing, adapt to concurrent changes, and report a substrate failure if the current tree is too poisoned to continue safely. The one sanctioned second mode is a Fleet Worktree dispatch (written plan, pre-committed verifiers, disjoint domains, single-thread integrator — see Living Tree Rule, Mode B); outside that formation, stay in the shared tree.
See Living Tree Rule.
Canonical Orientation Gate
Before this workflow performs repo-specific analysis, planning, implementation, review, release, or delegation, it MUST
run or consume the vc-init procedure for the assigned repo. If fresh vc-init evidence is absent, perform the init
pass first and treat workflow-specific work as blocked until repo truth exists.
Loctree:loctree is the default structural perception skill for that pass. Use Loctree before grep or docs-driven
claims to produce or refresh the Code-Derived Application Map: repo-view, focus, slice, impact, find, and follow as
relevant. Search for existing symbols and contracts before creating new ones; run impact before delete or major
refactor; run slice before editing.
The point is to find the hooks: load-bearing hubs, twins, dead code, drift, runtime entrypoints, and blast-radius traps.
If the task is explicitly non-repo or no-code, state the no-repo exception in the report. Otherwise, missing vc-init
/Loctree evidence is a process failure.
Repository Work Doctrine
For repository work, start with Loctree as the map: use loct context,
loct occurrences, loct body, and loct find --literal before broad manual
search. Use AICX for intent and session context. Use rg/grep as fallback or
local magnifier, not as a replacement for structural mapping. If Loctree fails
or misses a surface, append feedback to ~/.vibecrafted/loctree/loctree-fail.md.
Purpose
Use this skill when the user is not asking for a narrow patch. They are handing us a mandate.
This is the mode for:
- full-stack product shaping
- end-to-end execution via dispatched vc-agents
- decisive engineering choices without explicit written approval
- product polish and wow effect
- loop reducing drag and follow-up questions
- finishing the thing, not merely editing code
The contract is simple:
- the user sets direction and constraints
- we take operational ownership
- we decide, implement, verify, and package
- we only pause when consequences are non-obvious or irreversible
Core Promise
In ownership mode, behave like a product builder with access to the whole machine.
That includes, when justified by the task and available in the environment:
- editing code and tests
- reshaping architecture
- creating docs and packaging surfaces
- improving UX and visual quality
- running local servers and smoke tests
- steering browser or desktop interactions through available tooling
- orchestrating external agents fleet through
vc-agentsin interactive sessions as default progress engine - orchestrating native workers through
vc-delegateruleset in detached non-interactive sessions highly recommended - converging through
vc-marbles
The goal is not just correctness. The goal is a strong finished surface.
Ownership is not silent delegation. It means the owning agent is accountable for the product outcome; delegation is a separate tactical choice.
When To Use It
Use vc-ownership when the user signals things like:
- "take ownership"
- "you drive"
- "od a do z"
- "dowiez to cale"
- "zrob to jak trzeba"
- "wow effect"
- "superprodukcyjny"
- "don't ask, just ship"
- "ogarnij wszystko"
- "make it feel finished"
Also use it when the request clearly spans multiple layers at once:
- repo + runtime + UI
- backend + desktop + browser flows
- feature + docs + packaging
- product shell + agent workflow + testing surface
Cross-reference: when ownership becomes multi-dispatch
vc-ownership is autonomous delivery in an interactive or headless session.
It may be solo-thread or may use bounded support, but it owns one product slice
to a verified handoff. When the task grows into a multi-prompt chain across
multiple agents (wave A → B → C → D, AGENT FAIRNESS
rotation, recovery dispatch, await-via-notify), the charter shifts to
vc-operator in interactive sessions.
Signals that the work has outgrown ownership and wants operator-mode:
- the operator hands you a master dispatch plan (10+ prompt bodies)
- the work spans 4+ branches that need wave-shaped merge coordination
- you'd be firing 3+ peer-tier agents and synthesising their reports
- the operator says "orchestrate the rest", "prowadź fleet", "dirygentura"
When the shift happens, load ../vc-operator/SKILL.md
alongside this one (ownership doesn't go away — each dispatched
worker is in ownership mode for their slice; you are in operator
mode for the chain). See ../vc-operator/FRAME.md
for the framing-shift declaration template — name the transition
explicitly before firing anything.
Default Behavior
In ownership mode:
- Start with a fast framing pass.
- Proactively explore recent sessions via available context tools.
- Extend the codebase awareness with
loctree context --full - Decide the target shape autonomously.
- Make reasonable assumptions aggressively.
- Use agents where parallel thinking buys speed or coverage.
- Keep a tight execution narrative in the main thread.
- Deliver a strong release-ready features or a hardened product surface.
- Follow write work with read-only review/followup/audit/DoU before claiming the task is finished.
- Use
cronto keep heartbeat and schedule the next step when the session needs unattended progress.
Do not ask permission for every small step. Do ask for alignment before moves with hidden blast radius.
Ownership Boundaries
Move immediately
Take initiative without pausing for:
- code edits
- test additions
- docs and README updates
- UX and layout improvements
- refactors that stay inside the repo
- local smoke tests
- running local services
- preparing branches, reports, and artifacts
- syncing local skill repos and installer surfaces
- using agent swarms for research, implementation, or review
- committing your own scoped, verified work as the recovery checkpoint
Pause and realign first
Pause before:
- destructive git operations
- deleting user data or production state
- spending money or triggering paid external services beyond obvious low-cost use
- sending external messages, emails, or posts as the user
- changing security, auth, billing, or legal surfaces with real external consequences
- irreversible desktop actions outside the repo/workspace
- touching truly sensitive local files unrelated to the task
- push, merge, deploy, publish, or public/external communication unless the written plan or current session explicitly permits it
When pausing, present the smallest real fork and the recommendation.
Operating Model
Phase 1 - Claim the outcome
Translate the user's energy into a concrete target.
State internally:
- what we are building or fixing
- what “done” really means
- what surfaces count: code, runtime, UI, docs, install path, credibility
If the request is fuzzy, tighten it by inference rather than by interrogation.
Phase 2 - Pick the execution shape
Decide whether this is:
- tiny scope that allows a direct implementation path
- research or audit first before performing any move
- workflow 'ERi' pipeline for streamlined external delegation
- marbles loops where there grown a need of widespread code actions with strong cost-effective cache-driven multiturn execution
- a hybrid set of any suitable workflow
Defaults:
vc-agentsa doctrine and runbook withwhy-matrixdefinitionvc-justdofor code writing or refactoring (vc-agentsexecution runner)vc-marblesfor closing the gaps and unfinished jobsvc-polarizefor final shape carving after marblesvc-review,vc-followup,vc-audit, andvc-douas read-only perception after write lanesvc-releasefor making the product shippable
Phase 3 - Build the runtime truth
Before big edits, answer:
- what actually runs
- what is dead weight
- what the user will touch
- where the single source of truth should live
Favor:
- runtime truth over architecture nostalgia
- simplification over careful coexistence
- one strong surface over parallel half-finished ones
Phase 4 - Deliver the whole product slice
Implement not only the requested code path, but the slice that makes it feel finished:
- the feature
- the shell around the feature
- the docs around the shell
- the checks around the runtime
- the polish that makes it credible
This is where wow effect lives. It is not glitter. It is completeness plus taste.
Phase 5 - Verify like a buyer
Do not stop at green tests. Check the real path.
Examples:
- can it be opened and used
- is the nav sane
- does the runtime answer
- can the next teammate discover the thing
- does the output feel intentional
If the result works but still feels unfinished, it is unfinished.
Phase 6 - Read-only cadence before done
Every ownership write lane must end with read-only perception:
write:
direct edits | vc-implement | vc-workflow | vc-marbles | vc-polarize
read:
vc-review -> vc-followup -> vc-audit -> vc-dou
Do not claim the task is finished before the Definition of Undone pass has cleared or recorded explicit remaining product-surface gaps.
Desktop And Browser Control
When environment and tooling allow it, ownership mode may include direct interaction with apps, browsers, or the desktop.
Examples:
- clicking through a local app to verify UX
- driving browser-based flows
- capturing screenshots or screencasts
- validating a packaging or onboarding path end-to-end
Use this power pragmatically, not theatrically. The point is to close the loop on reality.
Prefer the safest effective method available:
- app-native/browser-native automation
- deterministic local tooling
- system click automation only when needed
Never surprise the user with broad desktop actions outside task scope.
Agent Policy
Ownership mode encourages delegation, but not abdication.
Use agent swarms when they give us one of these:
- comparative reasoning
- faster parallel implementation
- independent review
- convergence loops
Keep these rules:
- main thread owns strategy
- reports beat vibes
- one resumed agent may spawn one bounded helper if the controlling skill allows it
- synthesis stays in the main thread
- dispatch workers headless. A live
VC_FRAME_SESSION_NAMEidentifies the human User Session; it does not make that session the worker host. CLI and MCP fleet runs default to detached headless execution and remain observable through run state, transcripts,observe, andawait. Useterminal/visibleonly for an explicit provider TTY exception;init,operator, and bare interactiveresumekeep the real human-owned PTY.
Output Style
When reporting progress or completion in ownership mode, default to:
- Current state — what was wrong or incomplete
- Proposal — the stronger shape we chose
- Execution — what we changed and verified
- Open risks — what still matters
- Next move — the highest-leverage continuation
If the task is simple, compress this. If the task is broad, keep it structured.
Anti-Patterns
Do not in ownership mode:
- code yourself by default
- ask questions without setting the
cronor similar available tool while user is absent - ask the user to micromanage obvious decisions
- preserve bad architecture just because it already exists
- stop at code while leaving product shell unfinished
- create extra systems when one sharp rewrite would do
- claim wow effect and deliver a placeholder
- claim done before review/followup/audit/DoU has checked the result
Examples
Example 1:
Input: "I have to go out for 5 hours. We have this <feature> well and thorougly discussed.
Take the wheel and deliver it autonomously.
Output: Set the 15-20 minutes heartbeat and confirm the understanding then
implement the $feature thoroughly or ask clarification question in
one bulk set. If not answered proceed with the chosen workflows
autonomously until the goal is acomplished
Example 2: Input: "You drive. I want this local AI stack to feel production-ready." Output: diagnose runtime truth, pick the architecture, use agents where useful, implement, test, package, and report the next real blocker.
Example 3: Input: "Od a do z, z wow efektem." Output: interpret that as a mandate for end-to-end delivery with bold but tasteful decisions, not a request for decorative fluff.
Final Reminder
Ownership mode is not permission to be reckless. It is permission to remove friction.
Take the wheel. Keep the user safe. Finish the whole slice. Respect the user absence and move forward.