platform.gitlab.review-finding-schema
Use to produce structured GitLab code review findings and the final GitLab review result.
Best use case
platform.gitlab.review-finding-schema is best used when you need a repeatable AI agent workflow instead of a one-off prompt.
Use to produce structured GitLab code review findings and the final GitLab review result.
Teams using platform.gitlab.review-finding-schema should expect a more consistent output, faster repeated execution, less prompt rewriting.
When to use this skill
- You want a reusable workflow that can be run more than once with consistent structure.
When not to use this skill
- You only need a quick one-off answer and do not need a reusable workflow.
- You cannot install or maintain the underlying files, dependencies, or repository context.
Installation
Claude Code / Cursor / Codex
Manual Installation
- Download SKILL.md from GitHub
- Place it in
.claude/skills/review-finding-schema/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How platform.gitlab.review-finding-schema Compares
| Feature / Agent | platform.gitlab.review-finding-schema | Standard Approach |
|---|---|---|
| Platform Support | Not specified | Limited / Varies |
| Context Awareness | High | Baseline |
| Installation Complexity | Unknown | N/A |
Frequently Asked Questions
What does this skill do?
Use to produce structured GitLab code review findings and the final GitLab review result.
Where can I find the source code?
You can find the source code on GitHub using the link provided at the top of the page.
SKILL.md Source
# Review Finding Schema
Use this skill whenever a GitLab review agent or PM coordinator emits findings.
## Final Result Contract
The PM coordinator final answer must be exactly one fenced JSON block. The first content line inside the fence must be `GITLAB_REVIEW_RESULT:`.
```json
GITLAB_REVIEW_RESULT:
{
"stage": "closed",
"status": "ok",
"summary": "Concise review conclusion grounded in the supplied diff.",
"findings": [],
"nextActions": []
}
```
Required fields:
- `stage`: always `closed` for the final PM result.
- `status`: one of `ok`, `blocked`, `failed`.
- `summary`: short human-readable conclusion.
- `findings`: array of finding objects, empty when no concrete issues are found.
- `nextActions`: array of short strings, empty when no follow-up is needed.
Do not wrap the result in another object. Do not add Markdown prose outside the JSON fence.
## Finding Object
```json
{
"title": "Short finding title",
"body": "Evidence, impact, and suggested change.",
"severity": "major",
"category": "correctness",
"file": "src/example.ts",
"oldLine": 12,
"newLine": 18,
"suggestion": {
"replacement": "return validate(input)",
"confidence": "high"
},
"source": "pm-coordinator"
}
```
Required finding fields:
- `title`
- `body`
- `severity`
Optional finding fields:
- `category`
- `file`
- `oldLine`
- `newLine`
- `suggestion`
- `source`
Allowed severities: `info`, `minor`, `major`, `critical`, `blocker`.
Allowed categories: `correctness`, `security`, `testing`, `performance`, `maintainability`, `frontend`, `architecture`, `docs`, `config`.
## Suggestion Rules
Use `suggestion` only for a small, local replacement that can be applied to a changed diff line. The publisher may render it as a GitLab suggestion block only after inline position validation passes.
`suggestion` fields:
- `replacement`: replacement code only. Do not include markdown fences.
- `confidence`: `low`, `medium`, or `high`.
Do not include a suggestion when:
- the location is uncertain.
- the fix spans multiple files or multiple hunks.
- the fix needs product, security, or architecture approval.
- the replacement includes markdown fences.
When in doubt, put the recommendation in `body` instead of `suggestion`.
## Evidence Rules
Only include `file`, `oldLine`, or `newLine` when the location is grounded in the supplied GitLab diff evidence and review line map.
Never guess line numbers. Use the `Review line map for file/newLine/oldLine fields` rows when present:
- added lines use `newLine`.
- deleted lines use `oldLine`.
- unchanged context lines inside a hunk may use `newLine`; the publisher can map it to the matching `oldLine`.
If the exact diff hunk line is uncertain, omit line fields and let the publisher create a top-level summary finding.
Do not create findings for:
- style-only preferences
- generic best practices without diff evidence
- files skipped by filters
- behavior outside the supplied MR or commit diff
## Status Selection
Use `ok` when review completed, even if findings exist.
Use `blocked` when the diff is too large, truncated, overflowed, empty after filters, or otherwise lacks enough evidence to review.
Use `failed` when the review workflow itself failed and no reliable review conclusion can be produced.
## Minimal Valid Outputs
No findings:
```json
GITLAB_REVIEW_RESULT:
{
"stage": "closed",
"status": "ok",
"summary": "No concrete issues were found in the supplied diff.",
"findings": [],
"nextActions": []
}
```
Blocked:
```json
GITLAB_REVIEW_RESULT:
{
"stage": "closed",
"status": "blocked",
"summary": "Review was blocked because the supplied GitLab diff was truncated or too large.",
"findings": [],
"nextActions": ["Split the MR or request a manual review for the omitted diff."]
}
```Related Skills
platform.gitlab.verification-matrix
Use to plan minimal QA checks for GitLab review findings.
platform.gitlab.subagent-prompts.tech-architect
Prompt template for the GitLab review technical architecture custom subagent.
platform.gitlab.subagent-prompts.spec-writer
Prompt template for the GitLab review discovery and spec context custom subagent.
platform.gitlab.subagent-prompts.security-agent
Prompt template for the GitLab review security custom subagent.
platform.gitlab.subagent-prompts.risk-qa
Prompt template for the GitLab review QA and risk custom subagent.
platform.gitlab.subagent-prompts.frontend-designer
Prompt template for the GitLab review frontend design custom subagent.
platform.gitlab.subagent-prompts.developer
Prompt template for the GitLab review developer custom subagent.
platform.gitlab.subagent-prompts.auto-fixer
Prompt template for the GitLab review auto fixer custom subagent.
platform.gitlab.spec-gate-review
Use to decide whether a GitLab review has enough product and technical context to proceed.
platform.gitlab.security-review-policy
Use for security review of GitLab MR or commit diffs.
platform.gitlab.pm-risk-routing
Use to choose which custom subagents should be created for a GitLab review run.
platform.gitlab.gitlab-mr-review-workflow
Use for GitLab merge request review runs triggered by @Nine1bot comments or merge request webhooks.