Policy Fields
Configure how Critique reviews pull requests — strictness, specialists, escalations, and models.
Policies control how hard Critique reviews and which models run. Configure them in Dashboard → Automation at installation or repository scope.
Resolution order
- System defaults — Safe baseline when nothing is stored.
- Installation policy — Applies to all repos under that GitHub App install unless overridden.
- Repository policy — Overrides installation for one repo.
Most fields use repository-over-installation. Path escalations inherit from installation when the repository list is empty. Model fields inherit when the repository leaves a model unset.
enabledSpecialists
| Type | List of specialist kinds |
| Default | Security, Tests, Architecture, Performance |
Which parallel specialists run. Evidence gathering and synthesis always run.
Path escalations can add specialists at review time even if they are not in this list.
strictness
| Type | INFO, WARNING, or FAIL |
| Default | FAIL |
Controls how findings map to the GitHub check:
| Mode | Behavior |
|---|---|
INFO | Findings are advisory; check stays non-blocking |
WARNING | Findings surface as warnings; merge usually still allowed |
FAIL | FAIL-severity findings can fail the check and block merge when required |
Also flows into Remedy scope as review context.
rerunAllowed
| Type | Boolean |
| Default | true |
When false, users cannot re-trigger a review from the dashboard API (manual @critique /review may still apply depending on product settings).
pathEscalations
| Type | List of rules |
| Default | Auth and billing pattern bundles |
Each rule has:
- label — Human name (for example
auth-boundaries) - patterns — Substrings matched against changed file paths (case-insensitive)
- requireSpecialists — Specialists forced on when any path matches
Default bundles commonly escalate Security and Tests on auth paths, and add Architecture on billing-related paths.
Matched escalations appear in the run’s “why this ran” context in the dashboard.
reviewLeadModel
| Type | Model id or inherit |
| Default | Inherit system default |
Overrides the synthesis model. Must be allowed for your plan and the review-lead role.
reviewSpecialistModel
| Type | Model id or inherit |
| Default | Inherit system default |
When set, all specialists use this single model (per-specialist defaults are cleared).
remedyModel
| Type | Model id or inherit |
| Default | Inherit system default |
Overrides the model used for Remedy runs tied to this repo or installation.
Where to edit
Use the Automation UI for installation and per-repository policy editors. Changes apply to the next review run — they do not rewrite history on past runs.
Most common tweak
Teams that want findings without blocking merge often move strictness from FAIL to WARNING. Teams that want maximum signal before merge keep FAIL and tighten path escalations on sensitive directories.