Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Provide an explicitly requested, repository-grounded engineering review using contextual heuristics for code clarity, SOLID and design, testing, refactoring, architecture, and professional practice. Use only when the user invokes `$engineering-wisdom` or explicitly asks for this installed engineering-wisdom pack; do not turn its advice into repository policy or automatically rewrite an application architecture.
.claude/skills/hoangnb24-engineering-wisdom/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | 78% | 0% |
| case-02 | ✓→✗ | ▼ Worse | -83% | 0% |
| case-03 | ✓→✗ | ▼ Worse | -66% | 0% |
| case-05 | ✓→✗ | ▼ Worse | -38% | 0% |
| case-11 | ✓→✗ | ▼ Worse | -46% | 0% |
Give practical advice without treating a school of thought as universal law. Repository intent, code, tests, and observed behavior remain authoritative.
AGENTS.md, docs/WORKFLOW.md, and only the repository materialrelevant to the request.
not activation.
Keep reviews read-only. For changes, follow repository authority and validation rules before editing.
heuristics that match observed evidence. Read references/sources.md when attribution or the intellectual basis matters.
external input, adapter semantics, shipped artifact, and cumulative state that are relevant. Do not assume isolated core tests prove those boundaries.
For each material finding, keep these fields separate:
command result, or behavior.
the heuristic would make the system worse.
None, or a separatelylabeled proposal naming the repository owner, authority still required, possible mechanism, and removal condition. Never describe advice as an existing rule unless repository evidence already establishes it.
confirm or falsify the recommendation. Match proof to the claim: test pure policy in isolation, startup at the composition root, and delivery through the artifact users actually load.
Rank findings by user impact and change risk. Prefer one concrete example over several abstract claims.
a different shape.
targets, or dependency rules without repository-owned authority.
dependency when its local benefit outweighs the heuristic.
If the user requests implementation and authority is sufficient, make the smallest coherent change and run the repository's own proof. Keep the review rationale separate from the evidence that the change works.
Start with the outcome in one sentence. Then report only material findings:
textFinding: <specific effect> Observation: <repository evidence> Heuristic: <contextual rule of thumb and applicability> Trade-off: <counter-pressure or harm condition> Proposed repository-owned enforcement: None | <owner, missing authority, mechanism, removal condition> Verification: <command, test, measurement, or observation>
End with unresolved risks or None. Do not add a policy checklist by default.
Other measured skills in the registry, with their headline benchmark lift.