| name | archi-advice |
| description | Produce architecture improvement advice from existing-code Archi analysis. Use when the user wants concrete modification suggestions, sequencing advice, or a practical refactor roadmap based on the current codebase rather than task-goal planning. |
archi-advice
This skill synthesizes existing-code Archi review outputs. It should combine
full analysis plus current diff context when relevant before giving
architecture modification advice.
Trigger
- Concrete architecture improvement plan
- Refactor roadmap
- Existing-code redesign direction
- Sequenced architecture recommendations
- "What should we do next?"
Preferred Workflow
- Inspect the local command shape:
archi --help
- Run baseline analysis. If help includes
--full, use:
archi --full
If help lacks --full, use:
archi .
- If the user is evaluating active changes, also run incremental review. If
help includes
--full, use:
archi
If help lacks --full but includes --diff, use:
archi --diff .
- Do not run goal, plan-review, or fix-advice commands. Synthesize advice from
the baseline output, incremental output, and the current codebase.
Read Outputs
.architec/architec-summary.md
.architec/architec-analysis.json
Use the full code review as the primary frame. Use diff output as a modifier
when the user is asking about current changes.
Output Rule
- Lead with the current architectural position and score.
- Give a prioritized improvement plan, not just observations.
- Prefer a phased plan: immediate, next, later.
- Advice should be grounded in code-review findings, then adjusted by diff
context when relevant.
- Do not paste raw JSON.
Avoid Misuse
- Do not use this skill without a baseline full analysis.
- Do not produce a refactor roadmap from diff context alone; use it only after
archi-full establishes the structural baseline.
Example Prompts
- "Based on the current architecture, what should we improve next?"
- "Give me a concrete architecture refactor roadmap for this repo."
- "Turn the current Archi findings into a phased refactor plan."
Standard Output Template
Use this response shape:
Current Position
- baseline score: <score>
- current reading: <one-line interpretation>
Immediate
- <action 1>
- <action 2>
Next
- <action 1>
- <action 2>
Later
- <action 1>
- <action 2>