<< All versions
Skill v1.0.1
LLM-judged scan95/100ingvarconsulting/unica/code-review
+3 new
──Details
PublishedAugust 11, 2026 at 07:17 AM
Content Hashsha256:79a8d275a1117fba...
Git SHA9d2cfeb3a576
Bump Typepatch
──Files
Files (1 file, 2.6 KB)
SKILL.md2.6 KBactive
SKILL.md · 37 lines · 2.6 KB
version: "1.0.1" name: code-review description: "Код-ревью BSL и изменений 1С. Используй когда пользователь явно просит review, ревью diff/PR/модуля/изменения, поиск дефектов, регрессий, рисков или недостающих тестов."
Code Review
MCP routing
- Preferred path: use MCP
unicatoolsunica.code.search,unica.code.definition,unica.code.outline,unica.code.graph,unica.code.diagnostics,unica.meta.info,unica.standards.explain,unica.standards.search,unica.project.map, andunica.runtime.execute. - Use
unica.*.infotools before reviewing code that depends on metadata shape, form structure, rights, DCS, MXL, or interfaces. - Do not call internal analyzer, standards, runtime, or package adapters directly. They are hidden behind MCP
unica.
Review stance
Lead with findings. Order them by severity and ground each finding in a file/line reference, reproducible path, or diagnostic output. Keep summaries secondary.
Workflow
- Identify the review scope: changed files, target source-set, affected metadata objects, public entry points.
- Resolve changed exported methods and entry points with
unica.code.definition; inspect large modules withunica.code.outline. - Use
unica.meta.infofor affected metadata objects to connect the review scope with modules, roles, subscriptions, functional options, and predefined items. - Use
unica.code.graphfor callers, callees, neighbors, and impact analysis when a changed method/node can be resolved. Useunica.code.searchfor handlers, literals, query fragments, and non-method tokens. - Inspect metadata with
unica.*.infowhen code depends on object structure. - Run
unica.code.diagnosticswhen the review includes BSL code. Usemode=filefor touched modules andmode=workspaceonly when the review scope is broad. Useunica.standards.explainfor diagnostic codes or standards-sensitive claims. - Check high-risk 1C patterns: transaction boundaries, query-in-loop, server/client context, privileged mode, broad rights, background jobs, external calls, temporary files, and silent exception handling.
- Verify with
unica.runtime.executesyntax/tests when feasible; otherwise state the exact unverified risk.
Output
- Findings first: severity, path, issue, impact, suggested fix.
- Then open questions or assumptions.
- Then brief change/test summary only if useful.
Do not rewrite the code during a review unless the user explicitly asks for fixes after the review.