# Cisco & Stanford on Why Skills Are the New Code

## Executive summary

The industry is shifting from viewing software development around explicit code and implementation toward one centered on high-level intent and 'skills.' This paradigm requires a layered agent stack (models, tools, context, harnesses) that must be managed rigorously. Experts highlight that skill sprawl leads to failure through overlap, drift, and lack of activation visibility. Crucially, the consensus is that achieving business value relies less on deploying increasingly powerful frontier models and more on sophisticated context engineering and centralized management of skills.

## Key takeaways

- Skills as the New Code Paradigm: Software development is transforming from revolving around code/implementation to revolving around intent and instructions. Skills must be treated as first-class citizens, not just configuration files (Guy Podjarny).
- Three Failure Modes of Skill Sprawl: Skill sprawl negatively impacts teams through: 1) Overlap (multiple isolated implementations achieving the same outcome); 2) Drift (teams using outdated versions of skills); and 3) Lack of Activation (no visibility into whether a skill is actually being used by agents or humans).
- Context Engineering Beats Model Size: For achieving business value, smarter context engineering is more critical than deploying the most advanced model. Mid-tier models (e.g., Sonnet, GPT medium reasoning) are often sufficient when provided with proper context and structured skills.
- Instruction Following Leakage: Empirical testing involving 500 skills across 1,000 tasks revealed that over half (55%) of the time, models followed a skill's instructions even when the skill was not loaded. This suggests valuable information is already encoded in model weights.

## Technical details

- Layered Agent Stack Architecture: The agent stack is composed of primitives (Models) $ ightarrow$ Tools (giving models capabilities) $ ightarrow$ Context (guiding the model) $ ightarrow$ Harnesses (constraining or packaging components) $ ightarrow$ Factory Lines (composing all elements into full pipelines).
- Skill Management Metrics: Stanford research tested 500 skills, 1,000 tasks, and 19 permutations of models and harnesses to measure instruction following performance.
- Cost and Scalability Concerns: High-end model usage (e.g., Opus) can lead to rapidly increasing costs, especially when agents perform 'agentic fan out' by spawning multiple subagents.

## Practical implications

- Adopt a centralized registry and management system for skills to prevent overlap and ensure version control.
- Prioritize context engineering and structured data sharing over simply upgrading to the largest available LLM.
- Challenge teams to use mid-tier models (like Sonnet or GPT medium reasoning) as the baseline, reserving high-end models only for complex planning tasks.
- Implement robust activation tracking mechanisms to measure actual usage of developed skills.

## Topics

Agentic Systems, Context Engineering, LLM Architecture, Software Development Paradigms, Build Automation, Tessl, AI DevCon NYC 2026

Source: https://www.youtube.com/watch?v=PabKgxoegGg