What Is the Unit Economics of Commissioning a Custom Skill vs Building In-House?

The unit economics of commissioning a Claude Code skill versus building it in-house depend on two variables: how long it takes your team to build a production-grade skill, and what that time is worth. For most teams new to skill engineering, the answer surprises them.

TL;DR: The first skill your team builds takes 4 to 8 hours and produces a fair-weather skill. A commissioned production-grade skill takes 2 to 3 hours of an experienced AEM skill engineer's time and passes the bar check first run. Commission your first 3 to 5 skills; build in-house once the methodology is internalized.

What Are the True Costs of Building a Skill In-House?

The visible cost of building in-house is the developer hours spent. The invisible cost is the iteration overhead: how many hours of testing, fixing, and reworking a skill that doesn't meet the production bar. Both are real costs that compound over the first three months of a skill's lifecycle.

"The failure mode isn't that the model is bad at the task — it's that the task wasn't specified tightly enough. Almost every production failure traces back to an ambiguous instruction." — Simon Willison, creator of Datasette and llm CLI (2024)

In our commissions, we've tracked the in-house build attempt pattern across 23 clients who came to us after building their own first skills. The average first-build attempt by a developer new to skill engineering takes 4 to 8 hours and produces a fair-weather skill: one that works on easy inputs and breaks on edge cases. The same skill built by an experienced skill engineer takes 2 to 3 hours and passes the bar check on the first run.

The full cost breakdown for an in-house beginner build (intermediate complexity):

The U.S. Bureau of Labor Statistics reports software developers earned a median annual wage of $133,080 in May 2024 (BLS, OES 2024), which translates to roughly $64 per hour before benefits and overhead. Fully loaded rates of $75 to $150 are standard across US teams. This range is wide because skill quality varies significantly with skill engineering experience. A developer who has built 10 skills builds the 11th faster and better. A developer building their first skill pays the learning curve in time.

What Are the Costs of Commissioning a Skill?

Commission pricing for Claude Code skills runs $150 to $400 for intermediate-complexity skills, covering requirements extraction, SKILL.md authoring, output contract definition, reference file architecture, and testing across representative inputs (AEM pricing, 2026). That flat fee compares to developer rates of $75 to $150 fully loaded.

A first in-house build costs $300 to $1,200 before edge cases surface. The cost breakdown for a commissioned skill:

For a detailed breakdown of what commission pricing covers and what complexity tiers cost, see how much it costs to have a custom Claude Code skill built.

The commissioned skill arrives tested, with an output contract, a reference file architecture where applicable, and documented trigger conditions. The in-house build, unless the team has established a rigorous skill engineering process, typically lacks one or more of these components on the first version.

What Does the ROI Calculation Look Like for Each Approach?

The ROI comparison is the commission or build cost divided into the ongoing time savings the skill generates. A 2023 controlled study found developers completed tasks 55.8% faster when using AI assistance (Peng et al., arXiv 2302.06590, 2023). Skills encode that assistance as reusable infrastructure: the time saving is permanent, not dependent on prompt quality each session.

For a skill that saves one developer 20 minutes per day, 5 days per week:

• Weekly savings: 100 minutes
• Monthly savings: 400 minutes (6.7 hours)
• At a developer rate of $100/hour: $670/month in recovered time

At those savings, both a $350 commission and a $900 in-house build pay back inside 2 months. A McKinsey study of 40+ developers found code generation tasks completed 35 to 45 percent faster with generative AI tools (McKinsey, 2023). The ROI math is similar for both paths. The difference is in quality and time-to-value. IEEE Computer Society research puts software maintenance at 60 to 80 percent of total lifecycle cost (IEEE Computer Society, Software Maintenance Research, 2002). An in-house skill with no documented output contract or reference file architecture will carry that maintenance overhead in full. A commissioned skill with a defined contract and tested edge cases carries significantly less.

The commissioned skill is in production next week. The in-house build might reach production in the same timeframe if the developer has capacity. It reaches the same quality level in month 2 or 3, after edge cases have been identified and fixed.

For the detailed ROI framework, how to calculate the ROI of a Claude Code skill covers the full calculation including the quality multiplier.

When Does Commissioning Beat Building In-House?

Commissioning wins when your team lacks the methodology, the time, or the appetite for risk on a production-critical skill. Each skill-building session your developer cannot complete cleanly adds recovery overhead on top of the build hours. That overhead is not recoverable.

Research by Gloria Mark at UC Irvine found it takes 23 minutes to regain focus after an interruption (Mark, UCI, 2008). Commissioning wins in four scenarios:

1. The team has no skill engineering experience. The first skill pays the full learning curve. Commissioning it transfers that cost to someone who has already paid the curve.
2. Speed to production matters. A commission typically delivers in 48 to 72 hours. An in-house build with an unfamiliar developer and competing priorities takes 2 to 4 weeks.
3. The skill is high-stakes. A skill used in a production workflow, a client-facing output, or a compliance-adjacent process needs to pass the bar check before deployment. Commissioning provides that guarantee. In-house builds provide it inconsistently.
4. Developer time is the constraint. If the development team is at capacity, the opportunity cost of in-house skill building is the feature work not being done. Commission the skill, ship the feature.

The build-versus-commission decision framework covers this analysis in full.

When Does Building In-House Beat Commissioning?

In-house wins when the team has internalized skill engineering methodology, has seen 3 to 5 production-grade examples to calibrate against, and has a developer with capacity to build without competing deadline pressure. At that point, the learning curve cost is already paid and the incremental build cost drops below the commission fee.

After a team has commissioned 3 to 5 skills and studied the output, they have concrete examples of what production-grade looks like: how the output contract is structured, how descriptions are written, how reference files are organized. Building in-house from that baseline is faster and more likely to hit the bar on the first attempt.

Building in-house also wins for skills with deep internal context. A skill that encodes 3 years of institutional knowledge about a specific codebase requires 4 hours of knowledge transfer regardless of who builds it. If that transfer stays in-house, the resulting skill is often more accurate than one built by an external skill engineer who needs to reconstruct the context from documentation and interviews. The 2024 Stack Overflow Developer Survey found 62% of developers now actively use AI tools in their workflow (Stack Overflow, 2024). Teams at that adoption level have the internal context to specify skills that outside engineers cannot easily replicate. Even so, the Standish Group CHAOS Report found 50% of software projects exceed their original budget and deadline without delivering full scope (Standish Group, 2020). In-house skill builds are smaller in scope than enterprise software, but the failure mode is the same: underestimated edge cases, competing priorities, and no defined acceptance criteria. A commissioned skill sidesteps all three by design.

The hidden cost of NOT building skills is covered separately in the repeated context tax analysis. Whether you commission or build in-house, the tax for not having the skill is the same.

Frequently Asked Questions

For most teams, commissioning beats building in-house for the first 3 to 5 skills because the methodology transfer cost is zero: you receive a production-grade skill and a concrete reference architecture. In-house builds become the better option once the team has that reference and a developer with unblocked time to apply it.

Can I commission a skill and then modify it in-house afterward? Yes, and this is the recommended hybrid approach. Commission the initial build to establish the production bar, then maintain and extend it in-house. The commissioned skill provides a reference architecture for internal development.

What's the risk of commissioning a skill that doesn't work for my specific use case? The risk is low when the requirements brief is specific. The brief is the single biggest determinant of commission quality. A commission built on a vague brief produces a capable but generic skill. A commission built on a precise brief with 5 concrete input examples and a clear output contract produces a production-grade skill.

Is there a skill complexity threshold where in-house always wins? For very high-complexity skills requiring deep domain expertise or extensive reference file architecture, in-house builds with a domain expert on the team produce the best results. The threshold is roughly skills requiring more than 6 reference files and 3+ hours of domain expert time to specify correctly.

How do I evaluate the quality of a commissioned skill before accepting it? Test it against 10 representative inputs from your actual workflow, including 3 edge cases. A production-grade skill handles all 10. If it fails on more than 1 of the 10, flag the failure with the specific input and expected output and request a revision.

Will the cost of skill engineering go down as AI tools improve? The cost of producing a mediocre skill will go down. The cost of producing a production-grade skill will stay relatively constant because the value is in the specification and testing, not the writing. Specification quality and test coverage are the constraints, and those require human judgment on the current trajectory.

Last updated: 2026-04-28