Agent rollout decision record

Cursor Background Agents

Conditional pilot Use isolated background tasks in non-production repos; review generated branches/PRs and verify remote environment, secrets, and audit logs.

Check this agent for your environment Compare alternatives Methodology

Verdict

Conditional pilot

Background coding work can be piloted after remote-environment isolation and branch-review workflow are verified.

Decision depth: AI-assisted launch evidence record

Evidence audit: AI-assisted evidence audit

Permission footprint

Repository readinferred

File editsinferred

Shell commandsnot evidenced

Pull requestsnot evidenced

Issue/PR commentsnot evidenced

Network/API callsnot evidenced

Secrets/env varsnot evidenced

Sandboxingnot evidenced

Permission controlsnot evidenced

Audit loggingnot evidenced

Enterprise adminnot evidenced

Branch isolationinferred

Human approvalinferred

Production deploymentnot evidenced

Power level

Level 3: executes actions

Highest-risk evidenced power: pull-request creation

Controls separated

Required before pilot — organization must implement

least-privilege permissions, isolated branch, protected branches, mandatory human review, sandbox, audit logs

Vendor-native control evidenced

sandbox, branch isolation, security docs

Vendor-native control not evidenced

native sandbox, permission system, audit logs, admin console, SSO, data-retention settings

Pilot plan

Task class: IDE/background coding agent

Environment: non-production repository

Maximum access: pull-request creation

Reviewer: engineering manager and code owner

Success criteria: branch output is reviewable, tests pass, and no protected branch bypass occurs

Stop conditions: unscoped repo access or unclear remote execution model

Prohibited: direct merges, production credentials, deployment without approval

Blocking issues and unknowns

Blockers: runs commands is not explicitly evidenced in the current source set.

Unknowns: Remote execution and retention details require workspace-specific verification.

Limitations: IDE/team configuration changes risk.

Claim-level evidence

edits files

Cursor Background Agents: edits files

Status: reasonably-inferred · Reviewer: AI-assisted evidence audit · Confidence: 62/100

Source: Cursor Docs (official-docs)

Excerpt: Inspect diffs, run checks, and catch problems before you merge

Limitations: This evidence supports an inference; do not display as simply supported.

supports human approval controls

Cursor Background Agents: supports human approval controls

Status: reasonably-inferred · Reviewer: AI-assisted evidence audit · Confidence: 62/100

Source: Cursor Docs (official-docs)

Excerpt: Inspect diffs, run checks, and catch problems before you merge

Limitations: This evidence supports an inference; do not display as simply supported.

supports branch isolation

Cursor Background Agents: supports branch isolation

Status: reasonably-inferred · Reviewer: AI-assisted evidence audit · Confidence: 62/100

Source: Cursor Docs (official-docs)

Excerpt: Inspect diffs, run checks, and catch problems before you merge

Limitations: This evidence supports an inference; do not display as simply supported.

reads repository

Cursor Background Agents: reads repository

Status: reasonably-inferred · Reviewer: AI-assisted evidence audit · Confidence: 62/100

Source: Cursor Docs (official-docs)

Excerpt: Inspect diffs, run checks, and catch problems before you merge

Limitations: This evidence supports an inference; do not display as simply supported.