Install from the repository marketplace
codex plugin marketplace add mzored/SkipHow
codex plugin add skiphow@skiphowAdaptive virtual CTO · Claude Code + Codex
Own the product.
Let the agent own the engineering.
Describe what should become true. SkipHow makes one coding agent accountable as your virtual CTO: product decisions and protected actions stay with you; the agent owns research, architecture, planning, implementation, review, integration, and fresh proof.
One accountable CTO, optional workflows, and focused guidance used as needed.
01 · How it works
One skill between your request and a verified result.
SkipHow supplies the orchestration policy. The host supplies the model and execution. The boxes below are responsibility layers, not mandatory development stages.
-
You
Outcome and constraints
Describe the product result, the product choices that matter, and any protected action you are willing to take.
-
SkipHow
Authority and method policy
One CTO kernel keeps technical responsibility, the decision boundary, and completion rules present, then points to focused playbooks.
-
Host agent
Reasoning and execution
Claude Code or Codex chooses the method and uses its tools, sessions, worktrees, or subagents as needed.
-
Result
Evidence and uncertainty
The agent returns what changed, the evidence behind it, and anything that remains blocked or unverified.
02 · What changes
Stop being the agent's process manager.
- Delegate an outcome, not a procedure
- The agent chooses commands, methods, tests, branches, and decomposition.
- Keep product control
- Visible behavior, scope, cost, risk, privacy, rollout, and protected actions remain yours.
- Use process in proportion
- A small request stays small. Larger work can add research, tracking, review, or delegation without making them universal stages. The agent also keeps your system small, with fewer tests and tools, no leftover shims, and guards for money and security kept.
- Make "done" require evidence
- A plausible result is not completion. The report names the checks that ran and the uncertainty that remains.
- Use one working relationship
- The same entry covers questions, decisions, research, bugs, changes, review, triage, delivery, pause, and resume.
- Keep coordination inside the engineering work
- Continuity, delegation, reconciliation, integration, and any tracking your project calls for do not become the owner's workflow.
The promise is less manual supervision, not infallibility. SkipHow does not make the model smarter or guarantee that every instruction will execute.
03 · The contract
Give the agent autonomy without giving it product authority.
Product owner
Outcome and visible behavior
Scope, priority, cost, risk, privacy, rollout
Protected actions and genuine product choices
Coding agent
Research and technical design
Libraries, schemas, code, tests, branches
Project procedures and verified completion
SkipHow · decision, authority, and completion contract
No technical review required from the owner
The agent still follows every review, security, release, and delivery procedure required by the repository. SkipHow removes those engineering mechanics from the owner's role, not from the project.
04 · Why this design
I wanted the rigor without another method to operate.
Before SkipHow, I used GSD, OpenSpec, Superpowers, Matt Pocock's skills, BMAD, Paperclip, Mesa, and other agent systems on my own work. Each solved a real part of the problem: diagnosis, focused methods, specifications, task state, parallel work, or review.
I kept having to choose commands, approve technical artifacts, move work through phases, or remember which skill to invoke. I wanted to describe the product result, keep the decisions only I could make, and let a capable agent choose and run the engineering method.
05 · Choose honestly
Use the smallest layer that solves your problem.
- Use SkipHow
- You own the product outcome and want Claude Code or Codex to own the engineering method through a verified result.
- Use the base agent
- Your agent already maintains this authority boundary and verifies completion reliably. Another instruction layer adds little.
- Use a skill library
- You want to discover and invoke separate methods yourself.
- Use a workflow framework
- You want people to inspect and approve specifications, phases, tickets, or the development method.
- Use a runtime orchestrator
- You need persistent agent teams, queues, budgets, leases, scheduling, governance, or a control plane.
06 · Install
One skill. Two supported hosts.
Install the same shared skill package
claude plugin marketplace add https://github.com/mzored/SkipHow.git
claude plugin install skiphow@skiphowFor ordinary-language governance, ask the installed skill once to enable itself: $skiphow Enable SkipHow as my default virtual CTO on this machine. in Codex, or the same request through /skiphow:skiphow in Claude Code. It resolves the instruction file your host actually reads, previews the change, writes one reversible block after you confirm, and reports configured, available, and loaded as separate facts. This setup installs no hook and grants no action. Ordinary-language loading was observed once each on Codex with the exact 4.1.0 and 4.2.0 packages and has not been observed on Claude Code; no activation mode has a measured reliability. $skiphow and /skiphow:skiphow remain the explicit fallback. Updates and uninstall steps are in the owner guide; what each host has shown is in the dated support summary and the security policy.
07 · Evidence
Claims stop where the receipts stop.
Historical 2.x controlled runs showed product questions surfacing before dependent work, fully specified requests proceeding without engineering questions, diagnosis without weakening tests, and false-positive fixes being rejected.
Those are observations, not a reliability rate. Comparative advantage over a base agent, automatic selection reliability, and delegation beyond the act of dispatching a delegate remain Unverified.
08 · FAQ
Short answers before you install.
Do modern agents already do this?
They already do much of the engineering. SkipHow does not make them smarter; it makes ownership, authority, method selection, and evidence of completion explicit. If your base agent already holds that contract reliably, use it alone.
Does SkipHow orchestrate agents?
Yes, at the instruction level. It can guide the host agent to plan, choose methods, decompose, delegate, review, and reconcile work when the task calls for it. Each delegate is briefed to act as CTO of its own assignment: it fixes what it finds inside that assignment and returns the rest to the lead, who decides what happens to it. Reliable multi-agent delegation under that policy remains unverified.
Is SkipHow a standalone orchestrator?
No. Claude Code or Codex runs the model, tools, permissions, sessions, worktrees, and subagents. SkipHow provides no scheduler, queue, persistent worker service, budget enforcement, or control plane.
Who counts as a product owner?
Anyone who owns the outcome for this request: founder, product manager, designer, domain expert, or engineer. Technical fluency does not change the decision boundary.
Will it deploy or change access without asking?
Not beyond what you have already authorized. An established owner-authorized non-production workflow covers routine push, pull request, CI, and merge, and the agent checks that the destination and its actual effects are still covered. Production and live-data changes, public releases, credentials, access changes, repository settings, payments, and material deletion require applicable explicit authorization; staging and other isolated non-production environments are ordinary engineering unless your own rules say otherwise.