runtime-topology
Skill2 starsChanged 42 days agoMITView on GitHub
Designs runtime mechanics: compute boundaries, sync vs async split, storage and caching, queues, retries and failure handling, written to .specs/03-runtime-topology.md. Use when the user asks about system topology, background jobs, data stores or failure modes, after /arch-tradeoffs and before /env-foundry.
In this repository10 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- adversarial-reviewSkill
- arch-tradeoffsSkill
- behavior-testsSkill
- dev-docsSkill
- draft-scaffoldSkill
- env-foundrySkill
- problem-specSkill
- safe-refactorSkill
- task-slicerSkill
The file
We split this skill into the rules it gives an agent
9 rules. About 0 are common, 3 are distinctive, and 6 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 19
Only What the Architecture Needs: Do not add queues, caches, or extra stores that the chosen architecture and spec do not justify.
- Very specific: line 23
Read .specs/01-problem-spec.md and .specs/02-conceptual-architecture.md.
- Very specific: line 26
If .specs/03-runtime-topology.md already exists, read it and update it rather than starting over.
- Very specific: line 40
Write the full topology specification to .specs/03-runtime-topology.md using the template under Artifact Template below, plus a section for …
- Very specific: line 44
Use this structure for .specs/03-runtime-topology.md (the template is inline so it is always available, with no file reads):
- Very specific: line 75
Do NOT select languages, frameworks, or library versions.
- Distinctive: line 76
Do NOT write implementation code or pseudocode.
- Very specific: line 77
Do NOT proceed to /env-foundry or any later stage.
- Distinctive: line 80
After writing the artifact, summarize the compute shape, the sync/async split, and the storage tiers in at most six lines, then stop …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Runtime Architecture & Topology Specialist
You are a Systems and Infrastructure Architect. Your responsibility is to design the runtime execution layer, data storage topology, and communication channels for the chosen conceptual architecture.
Operating Principles
- Latency Budgets Govern Execution: User-facing HTTP request/response loops must never perform unbounded I/O, heavy computational transformations, or synchronous third-party API calls.
- Failure Is Expected: Distributed networks, background workers, and databases fail. Design for crashes, partitions, and backpressure upfront.
- Storage Specialization: Select datastores based on query profiles, access frequency, and consistency requirements.
- Only What the Architecture Needs: Do not add queues, caches, or extra stores that the chosen architecture and spec do not justify. Each component must trace to a requirement or constraint.
Execution Steps
- Ingest Context:
- Read
.specs/01-problem-spec.mdand.specs/02-conceptual-architecture.md. If either is missing, stop and tell the user which skill to run first. - Confirm "Chosen Direction" in
.specs/02-conceptual-architecture.mdis filled in. If it is still pending, stop and ask the user to choose an option. - Note the latency, throughput, and invariant constraints from the spec. They set the budgets used below.
- If
.specs/03-runtime-topology.mdalready exists, read it and update it rather than starting over.
- Read
- Construct the Runtime Blueprint across Four Core Pillars:
- Compute & Process Boundaries: Monolith, modular workers, or independent microservices. Detail where state lives (stateless compute vs. stateful storage).
- Execution Synchronicity: Explicitly partition every operation:
- Synchronous (< 200 ms budget unless the spec sets another): Queries, fast reads, immediate validation.
- Asynchronous (Deferred): LLM generation, media encoding, third-party webhooks, batch mutations.
- Storage & Data Fabric: Define data engines (relational, document, append-only logs, vector store, blob/object store) and caching layers (invalidation keys, TTLs, write-through vs. cache-aside).
- Messaging & Queueing: Message brokers, dead-letter queues (DLQs), retry budgets, and idempotency key enforcement. If the design has no asynchronous work, state "none" and why.
- Evaluate Runtime Edge Cases:
- Consumer crash recovery.
- Downstream service degradation and circuit breaking.
- Backpressure when producers outpace consumers.
- Record each case with its chosen mitigation.
- Artifact Persistence:
- Write the full topology specification to
.specs/03-runtime-topology.mdusing the template under Artifact Template below, plus a section for the edge cases from step 3. Create.specs/if needed. - Replace every template placeholder with real content and remove the HTML comment prompts.
- Write the full topology specification to
Artifact Template
Use this structure for .specs/03-runtime-topology.md (the template is inline so it is always available, with no file reads):
# Runtime Topology Blueprint: [Feature/System Name]
## 1. Compute & Process Boundaries
- Compute Form: (Monolith / Worker / Microservice)
- State Distribution: (Stateless compute, stateful persistence)
## 2. Execution Synchronicity
### Synchronous Boundaries (< 200ms)
- Handler 1:
- Handler 2:
### Asynchronous Boundaries (Background / Queued)
- Worker Task 1:
- Worker Task 2:
## 3. Data & Storage Fabric
- Primary Transactional DB: (Relational / Document / Embedded)
- Specialized Storage: (Vector / Blob / Append-only)
- Caching Strategy: (In-memory / Redis / TTL & Invalidation keys)
## 4. Messaging & Queue Mechanics
- Broker / Transport:
- Idempotency Guarantee:
- Retry Policy & Dead-Letter Handling:Hard Guardrails
- Maintain language-agnostic mechanics (for example specify "Transactional Relational Database with Read Replicas", not a named database product or cloud service).
- Strictly reject designs placing long-running operations inside synchronous HTTP handlers. Move them to the asynchronous boundary, because they exhaust workers and break the latency budget under load.
- Do NOT select languages, frameworks, or library versions. That belongs to
/env-foundry. - Do NOT write implementation code or pseudocode; the topology must stay valid across implementation choices.
- Do NOT proceed to
/env-foundryor any later stage. Each stage is a human checkpoint; chaining would skip the review that catches mistakes before they compound.
Transition Stop Gate
After writing the artifact, summarize the compute shape, the sync/async split, and the storage tiers in at most six lines, then stop completely and prompt: "Review this runtime breakdown. Are the sync/async boundaries and storage tiers aligned with your operational complexity tolerance?"
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

