Visual Studio Code

VS Code settings

Configure the released Cortex extension settings for agent behavior, validation, privacy, and style.

Open VS Code Settings and search for Pervaziv AI Cortex. Workspace settings can be committed only when they are appropriate for every contributor; keep machine-specific paths and local hook configuration out of shared settings.

Agent

Setting	Default	Purpose
`pervazivai-cortex.enableFileDecorations`	`true`	Mark vulnerable files in Explorer
`pervazivai-cortex.enablePersonalizedResponses`	`true`	Allow eligible cross-chat personalization
`pervazivai-cortex.autoApplyAgentEdits`	`true`	Apply edits after local safety checks; turn off to stage supported edits
`pervazivai-cortex.runPatchVerifier`	`false`	Run the advisory verifier after automatic validation
`pervazivai-cortex.shareVerifierTrainingFeedback`	`false`	Share redacted verifier feedback separately from chat memory
`pervazivai-cortex.actionSafety.showDetails`	`true`	Show policy reason codes for confirmations, blocks, and escalations
`pervazivai-cortex.githubTransactionsMode`	`disabled`	Choose `disabled`, `observe`, or `enforce` for governed GitHub writes
`pervazivai-cortex.globalAgentInstructions`	empty	Add coding expectations for every workspace
`pervazivai-cortex.relatedRepositories`	`[]`	Opt in repository IDs for cross-repository context
`pervazivai-cortex.inlineCompletions.enabled`	`false`	Enable preview inline completions
`pervazivai-cortex.nativeChat.enabled`	`true`	Expose the `@cortex` VS Code Chat participant
`pervazivai-cortex.nativeChat.includeActiveEditor`	`true`	Include bounded active-editor context with `@cortex`

Inline-completion debounce, timeout, prefix, and suffix limits are also available under pervazivai-cortex.inlineCompletions.*.

Skills and hooks

Setting	Default	Purpose
`pervazivai-cortex.hooks.enabled`	`true`	Allow configured local hooks during agent tool execution
`pervazivai-cortex.hookFiles`	`{}`	Set script files for `beforeCommand`, `afterCommand`, `afterFileEdit`, or `afterTask`
`pervazivai-cortex.inlineHooks`	`{}`	Set inline scripts for the same four lifecycle points
`pervazivai-cortex.skills.directories`	`[]`	Select skill directories that contain `SKILL.md`

Trust hook code before enabling it. Invalid paths are not sent, and filesystem or workspace roots are not valid skill directory entries.

Validation and mutation testing

Validation is off by default after agent edits. Configure default kinds from test, lint, typecheck, build, format, and static; the shipped default selection is typecheck, lint, and test. You can also set the timeout, changed-line coverage, Problems-panel publishing, and fix/verify attempt limit under pervazivai-cortex.validation.*.

Manual mutation testing is disabled by default. Its settings control local execution opt-in, engine (auto, stryker, or mutmut), timeout, maximum retained mutants, and whether a tests-only strengthening plan may be offered. Local execution is for trusted repositories and does not provide filesystem or network isolation.

Prompt, privacy, and style

Light prompt guardrails and quick privacy scanning are on by default.
Advanced and maximum prompt protection, deep chat scanning, and deep file scanning are opt-in and can require an eligible plan.
Style Profile is off by default. When enabled, it can set comment density, documentation format, naming, error handling, type strictness, function size, and test structure preferences.

Organization policy and server capabilities remain authoritative even when a local setting is enabled.