04-architecture
Skill1 starsChanged 53 days agoMITView on GitHub
Version: 1.0.0 Compatible: project_skills >= 1.0 Requires: discovery, project-context, business-architecture Outputs: architecture.md, architecture-gaps.md, decisions/ADR-XXX.md Understand, review, and document the project's architecture. This skill directs the agent to assess the existing architecture (if any), identify gaps, and make or record architectural decisions with reasoning. Every architectural decision needs a reason. Identify: - High-level components and their responsibilities - How components communicate (HTTP, queues, shared DB, events, in-process) - Data flow between components - Boundaries between subsystems - Deployment topology (processes, services,…
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 14
Every architectural decision needs a reason.
- Very specific: line 73
Every architectural decision that is not already documented must be recorded as an ADR.
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 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
04-Architecture Skill
Version: 1.0.0 Compatible: project_skills >= 1.0 Requires: discovery, project-context, business-architecture Outputs: architecture.md, architecture-gaps.md, decisions/ADR-XXX.md
Purpose
Understand, review, and document the project's architecture. This skill directs the agent to assess the existing architecture (if any), identify gaps, and make or record architectural decisions with reasoning.
Every architectural decision needs a reason.
Process
1. Map the current architecture
Identify:
- High-level components and their responsibilities
- How components communicate (HTTP, queues, shared DB, events, in-process)
- Data flow between components
- Boundaries between subsystems
- Deployment topology (processes, services, instances)
- External dependencies
2. Assess architectural style
Determine what style the project uses or should use:
- Modular monolith
- Microservices
- Serverless functions
- Single process
- Client-server
- Event-driven
- Layered
- Hexagonal / ports and adapters
- Other
Record why this style was chosen (or note "not documented" if unknown).
3. Evaluate against principles
For each component and interaction, ask:
- Is this the simplest thing that can work?
- Is this component's responsibility clear and focused?
- Are boundaries well-defined?
- Is the data flow understandable?
- Are dependencies justified?
- Is this architecture appropriate for the project's scale and team?
- Does it match the business domain?
4. Identify gaps
Record gaps in these categories:
- Missing documentation
- Unclear boundaries
- Hidden coupling
- Missing error handling strategy
- Missing observability
- Missing security boundaries
- Scaling bottlenecks
- Technology choices that no longer fit
- Missing backward compatibility strategy
- Missing migration strategy for data or schema
5. Make or record decisions
Every architectural decision that is not already documented must be recorded as an ADR. See templates/adr/.
An ADR must contain:
- Title
- Status (proposed, accepted, deprecated, superseded)
- Context: what problem this decision addresses
- Decision: what was decided
- Consequences: what this decision implies (positive and negative)
- Alternatives considered: what else was considered and why it was not chosen
Architecture Assessment Questions
- What is the simplest architecture that satisfies the business requirements?
- Are we over-engineering? Under-engineering?
- Does the architecture match the team's ability to maintain it?
- Can we deploy independently where it matters?
- Are there single points of failure?
- Is data integrity protected across component boundaries?
- Can we reason about the system? If not, why?
- Are there components that do too much? Too little?
- Are there circular dependencies?
- Is the dependency direction correct (dependencies point toward stability)?
Output
architecture.md
Architecture
============
Style: [architectural style]
Rationale: [why this style, or "not documented"]
Components:
- [name]: [responsibility], [communicates with: ...]
Data flow: [description or diagram]
Deployment: [how it runs]
External dependencies: [list]
Boundaries:
- [boundary description]architecture-gaps.md
Architecture Gaps
=================
[Gap ID]: [description]
Severity: [low | medium | high | critical]
Impact: [what this gap causes]
Recommendation: [what to do]decisions/
Each ADR as a separate file. Use templates/adr/ as the template.
In this repository17 other files it gives its agents.
- AGENTS.mdAGENTS.md
- llms.txtllms.txt
- 00-orchestratorSkill
- 01-discoverySkill
- 02-project-contextSkill
- 03-business-architectureSkill
- 05-documentationSkill
- 06-codebase-auditSkill
- 07-securitySkill
- 08-testingSkill
- 09-performanceSkill
- 10-auditSkill
- 11-controlled-source-of-truthSkill
- 12-context-engineeringSkill
- 13-phase-loop-deliverySkill
- 14-forensics-and-debuggingSkill
- 15-autonomous-loop-and-simplificationSkill
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.

