Skip to main content
A quick reference for the statuses, severity levels, and check results you’ll encounter in Oplane’s PR/MR reviews and threat models.

Requirement status

Each security requirement has a status indicating whether it has been addressed. The PR/MR comment shows each status as a colored circle:

Severity levels

Severity says how bad a threat is if nothing mitigates it, and guides your response. Oplane grades it when it assesses a requirement against your code, so a new requirement is Undetermined until then. The implementation status never changes it, so a Critical requirement you fix stays Critical and is marked Implemented. On threat model cards in a workspace, each severity with unresolved requirements (including Not applicable and Undetermined) appears as a compact chip showing how many are left to fix, next to a progress ring showing overall completion. Unreviewed requirements are grouped into a single “unreviewed” chip. Hover over the ring to see resolved/total counts per severity.

How severity is graded

Oplane grades every requirement on two axes, judged from the code as if the requirement’s control were missing. Impact is what this gap alone gives an attacker, and reachability is who can carry out the attack once the control is removed. A control Oplane finds in your code never lowers the severity. A perimeter that holds untrusted code counts as Signed-in external users. A secret committed to a repository is reachable by whoever can read that repository, and stored or logged data by whoever reads it. The severity is the cell where the two axes meet: When an axis is unclear, Oplane takes the higher value. If the data involved is credentials, payment instruments, health records, or another category of regulated personal data (names and addresses included), the severity goes up one level, to at most Critical. Fake or placeholder data doesn’t count, and a catastrophic impact never goes up, because that data is already part of it. A requirement that nobody can reach, or that’s irrelevant, is graded Not applicable instead. Each grade an agent makes comes with a note that states the impact, the reachability, whether the data raised the level, and the evidence in the code.

Who grades severity

Oplane’s pull request reviews and continuous threat models grade every requirement they assess. When you work through MCP, your AI assistant grades each requirement with update_security_requirement_severity, and the Oplane plugin skills do it as part of their workflow. To change a grade yourself, click the requirement’s severity in Oplane. Pick a level and the impact and reachability it rests on, pick Not applicable, or pick Reset to Undetermined to have the requirement graded again at its next assessment. Each change needs a note. Oplane’s reviews keep a severity a person set, unless the requirement’s impact or reachability has changed since.

Check status

Oplane posts an Oplane Security Review check on each PR/MR with one of these conclusions: By default, merge blocking is off and unresolved requirements report Neutral, which GitHub and GitLab treat as passing. To have the check report Action required when unresolved requirements reach a level you care about, turn on Block merges on unresolved findings per repository. Then pick a severity on the slider. GitHub branch protection treats Action required as non-passing, so it blocks the merge just like a failure. GitLab has no equivalent state, so the check reports a blocking Failed status instead. Combined with your branch protection rules, this acts as a merge gate. See Merge gating for the setup on GitHub and GitLab. The check’s summary counts unresolved requirements, for example 1 of 3 security requirements need attention, or All 3 security requirements resolved once everything is addressed. The check’s details link opens the generated threat model.
Info and Not applicable requirements never contribute to the failure threshold, even at the lowest Low setting. An unresolved Undetermined requirement blocks at every setting, because nobody has shown it sits below the threshold yet.

Comment structure

Oplane posts a single Oplane Security Review summary comment on your PR/MR and keeps it up to date as the review state changes. If the PR/MR reaches the limit on automatic reviews, Oplane posts a second, separate comment about the limit and leaves the summary comment unchanged. The summary comment contains:

Summary headline

States how many requirements need attention and links to the generated threat model.

Requirements table

Lists every requirement with its current status, title, and severity. Unresolved requirements are sorted first, ordered by severity (Critical → Info, then Not applicable and Undetermined). Each unresolved row includes fix links to hand the requirement to your coding agent.

Fix with Oplane button

Shown below the table when suggested fixes are turned on for the repository and at least one requirement is unresolved. It appears on the first review comment, even before Oplane has finished generating the fixes. It opens a page where you review the fixes and commit them to the branch. See Click to fix.

Agent instructions

An instruction block embedded in the comment’s markdown, hidden in the rendered view. Coding agents can pick it up to fetch the threat model and report implementation status via MCP. See MCP setup.

Respond to requirements

Learn how to respond to requirements and run local checks.