Agent Skills: Bridging AI and Senior Engineer Practices

Addy Osmani · article · 2026-10-08 source ↗ Reader 🖥️

Agent Harness Agent Skills Scope Discipline

AI coding agents are capable junior engineers with no instinct for the invisible, non-diff parts of senior engineering — specs, tests, reviews, scope discipline, verification — so Agent Skills exists to make those parts structurally non-negotiable, because they are no longer optional even when the 'engineer' is a model.

practical · argumentative — A short promotional post that nonetheless states a real thesis and four named principles, rather than pure sponsor content — treated per the 'short doc gets a short extraction' rule.

Structure

Part	Role
Framing: senior engineering is mostly invisible work	premise
Diagnosis: agents default to the shortest path to 'done' and skip that invisible work	evidence/diagnosis
Four core principles (process over prose, anti-rationalization tables, verification non-negotiable, scope discipline)	solution summary
Closing reframe: agents as capable juniors, not seniors, so the scaffolding isn't optional	conclusion

Key terms

capable junior engineer (applied to AI agents)
A capability/judgment framing for current AI coding agents: strong at execution, but — like a junior human engineer — lacking the instinct for the invisible, non-diff parts of the job (specs, tests, scope discipline) that experience teaches.

Propositions

explicitA senior engineer's job is mostly the parts that don't show up in the diff: specs, tests, reviews, scope discipline, and refusing to ship what can't be verified.conf: high
supportsAgent Skills · p2 — Restates the same 'invisible work defines seniority' premise that grounds the full essay's argument.
explicitdevelopedBy default, AI coding agents skip all of the invisible senior-engineer steps, taking the shortest path to 'done': asked for a feature, they write the feature, declare victory, and move on without a spec or a failing test first.conf: high
supportsAgent Skills · p1 — Identical diagnosis of the agent default-to-shortest-path failure mode, compressed into the promotional post.
extends ←Addy Osmani's Post (Agent Skills 0.6.3 release) · p1 — The earlier Agent Skills post names the problem (agents default-skip invisible senior-engineer steps); this release announcement is the direct follow-up investing specifically in the review stage of that same framework, naming review as the current bottleneck.
explicitAgent Skills is a system for bolting senior-engineer scaffolding back onto an AI agent by making the invisible, critical steps non-negotiable rather than optional best practice.conf: high
explicitScope discipline — touching only what was asked — is the single biggest determinant of whether an agent's PR is mergeable.conf: high
supportsAgent Skills · p9 — Both single out scope discipline as the single biggest determinant of a mergeable agent PR — same claim, same wording almost verbatim.
explicitAI agents are extremely capable junior engineers with no instinct for the invisible parts of the job, which means the senior-engineer parts of software work are no longer optional even when the engineer is a model.conf: high
extendsAgent Skills · p1 — The essay diagnoses the default-to-done failure mode; this post adds the sharper 'capable junior engineer, not senior' framing as the underlying reason the failure mode exists and recurs.
implicitFraming agent limitations as a seniority gap (capable junior vs. senior) implicitly assumes the right fix is porting existing human mentorship/review structures onto agents, rather than designing a fundamentally different governance model for non-human contributors.conf: medium

Arguments

Senior-engineer scaffolding (specs, tests, reviews, scope discipline, verification) must be structurally enforced on AI agents via a system like Agent Skills, not left as optional best practice.

Explicit premises:

AI agents default to the shortest path to 'done' and skip invisible disciplinary steps
Those disciplinary steps are what separates reliable software from code that breaks
Agents are capable junior engineers, and junior engineers' invisible-discipline gaps are exactly what senior scaffolding (mentorship, review, process) exists to cover

Implicit premises:

The junior/senior engineer metaphor transfers cleanly from human career development to model capability, i.e. an agent's gap is structurally the same kind of gap a junior human has, not a categorically different kind of limitation

Inferences:

If the gap is a seniority gap rather than a different-in-kind limitation, then the fix a team already has for junior humans (process, review, scaffolding) should in principle also work for agents — which is the premise the full essay (extract-agent-skills-osmani.json) builds out in detail

Questions addressed

✅ Why do AI coding agents produce code that looks done but isn't reliable?
They skip the invisible, non-diff work (specs, tests, reviews, scope discipline, verification) that senior engineers do by habit, because their default reward signal is 'task complete,' not 'task complete with evidence.'

Gaps

incompleteThe post frames the fix entirely as scaffolding bolted onto the individual agent (skills, workflows, anti-rationalization tables) and never addresses who performs the human senior-engineer role — review, judgment calls on ambiguous scope, final accountability — once a team scales up agent usage faster than it scales up senior human reviewers. A 'capable junior' needs a senior somewhere in the loop; the post doesn't say where that capacity comes from at team scale.
Read next: Look for team/org-level treatments of reviewer capacity and accountability once a large share of PRs are agent-authored — e.g. whether review throughput, not agent discipline, becomes the binding constraint.