“Once the skill is stable, look for one adjacent skill you could chain it with. Composition is where the leverage compounds.” — Lou
Session context: 2026-04-02_Mastermind — Lou described the architectural shift from monolithic prompts to composable skill modules. This sub-insight owns the composability facet of that original framing.
Core Idea
A skill built in isolation is useful. A skill built to chain with other skills is a force multiplier. The difference is intentional design: modular judgment units with clean interfaces can be sequenced, forked, and reused across contexts in ways a monolithic prompt can never be.
The verbatim passage from the original insight that anchors this facet:
“Once the skill is stable, look for one adjacent skill you could chain it with. Composition is where the leverage compounds.”
This sentence is brief, but it encodes a design philosophy. When you build a skill, ask not just “does this work?” but “what could this feed into, and what could feed into this?” A researcher skill that produces structured output becomes far more valuable when it feeds directly into an outliner skill, which feeds into a writer skill. Each unit does one job well; the pipeline does many jobs reliably.
Composability also distributes maintenance. When one skill in a pipeline needs updating — because a model changed, or your standards evolved — you update that unit without touching the others. A monolithic prompt that does research + outlining + writing requires a full rewrite when any one part needs adjustment. Modular skills require a targeted edit.
The shared-modes pattern (from Insight - Skill Composition — Many Small Skills With Shared Modes Beat One Big Monolith) extends this: a “mode” file shared across multiple skills in a pipeline ensures they all speak the same language — same format conventions, same audience assumptions, same quality standards — without duplicating that specification in every skill file. The mode is the shared contract that makes skills composable without renegotiating at every handoff.
Practical Application
When a skill is working well and you’re about to move on, pause and ask:
- What does this skill’s output look like? If it produces structured data or a document with a clear schema, it can feed another skill.
- What skill would naturally consume this output? Name it. Even if it doesn’t exist yet, naming the downstream consumer clarifies what format and fields your current skill should produce.
- What does this skill need to do its job? That’s the upstream skill. If you find yourself manually feeding the same type of input every time, the upstream is worth building.
- Can you define a mode? If you have two or three skills that should always produce output in the same register (same voice, same audience, same structure), extract that shared specification into a mode file and have each skill inherit it.
The leverage point: a library of 8 composable skills serves you better than 20 standalone skills of equal quality.
Related Insights
- Insight - Skill Composition — Many Small Skills With Shared Modes Beat One Big Monolith — the named pattern for composable skill architecture; the shared-modes technique is the practical implementation
- Insight - Design Tiny Tools for Maximum Composability — composability as a design principle from the tool-building angle; same concept at a different layer
- Insight - Design AI Systems for Maximum Composability and Minimum Context Pollution — system-level composability; the “minimum context pollution” clause is what makes chaining practical
- Insight - The Structured Object as Agent Memory — Agents Don’t Talk to Each Other, They Talk Through the Object — composability in multi-agent systems; the structured object is the inter-skill interface in an agent context
Evolution Across Sessions
Split from Insight - Skills Encode Judgment Into Persistent, Composable Intelligence (2026-04-08) when the hub crossed the 15-inbound threshold. This sub-insight owns the composability facet: what makes skills worth building beyond their individual utility — the network effects that emerge when they can be chained. For the persistence facet, see Insight - Skills Outlive Conversations — Persistence Is What Turns a Process Into Infrastructure. For the judgment-transfer mechanism, see Insight - Skills Are Judgment Transfer Vehicles — Not Just Reusable Prompts.