mirror of
https://github.com/stablyai/orca.git
synced 2026-10-04 08:02:09 +00:00
Brennan Benson
7f5141ae2d
* fix(structured-chat): deliver the permission posture through each transport's own contract Codex posture moves off app-server argv onto typed `thread/start` and `thread/resume` params. Manual states `on-request` / `workspace-write` explicitly instead of omitting the fields, which app-server resolved through the mirrored config.toml — a Manual thread on a home carrying `approval_policy = "never"` never prompted. Claude keeps its owned `--dangerously-skip-permissions` flag through SDK `extraArgs`; the SDK's typed bypass option emits a newer allow flag that older user-installed binaries reject. Posture is re-derived from current settings on every session acquisition. * fix(structured-chat): parse permission arguments as argv * fix(structured-chat): keep permission policy authoritative
76 lines
3.2 KiB
TypeScript
76 lines
3.2 KiB
TypeScript
import { describe, expect, it } from 'vitest'
|
|
import { codexStructuredPermissionPolicyForSettings } from './codex-structured-permission-policy'
|
|
|
|
const BYPASS = { approvalPolicy: 'never', sandbox: 'danger-full-access' }
|
|
// Approvals on, writes confined to the workspace. Verified against codex 0.153.4: both values are
|
|
// accepted on thread/start and thread/resume, and the reply echoes them back as the effective
|
|
// policy even when the home's config.toml asks for `never` / `danger-full-access`.
|
|
const MANUAL = { approvalPolicy: 'on-request', sandbox: 'workspace-write' }
|
|
|
|
describe('codexStructuredPermissionPolicyForSettings', () => {
|
|
it('bypasses when the user has never opened Agent settings', () => {
|
|
expect(codexStructuredPermissionPolicyForSettings({ agentDefaultArgs: {} })).toEqual(BYPASS)
|
|
expect(codexStructuredPermissionPolicyForSettings({})).toEqual(BYPASS)
|
|
expect(codexStructuredPermissionPolicyForSettings(null)).toEqual(BYPASS)
|
|
expect(
|
|
codexStructuredPermissionPolicyForSettings({ agentDefaultArgs: { claude: '' } })
|
|
).toEqual(BYPASS)
|
|
})
|
|
|
|
it('bypasses when Yolo wrote the flag, alone or beside other tokens', () => {
|
|
for (const codex of [
|
|
'--dangerously-bypass-approvals-and-sandbox',
|
|
'--dangerously-bypass-approvals-and-sandbox --model gpt-5.6-sol',
|
|
'--model gpt-5.6-sol --dangerously-bypass-approvals-and-sandbox'
|
|
]) {
|
|
expect(
|
|
codexStructuredPermissionPolicyForSettings({ agentDefaultArgs: { codex } }),
|
|
codex
|
|
).toEqual(BYPASS)
|
|
}
|
|
})
|
|
|
|
it('keeps quoted mentions and operands after -- in Manual', () => {
|
|
for (const codex of [
|
|
'--config "note=--dangerously-bypass-approvals-and-sandbox only as text"',
|
|
'-- --dangerously-bypass-approvals-and-sandbox'
|
|
]) {
|
|
expect(
|
|
codexStructuredPermissionPolicyForSettings({ agentDefaultArgs: { codex } }),
|
|
codex
|
|
).toEqual(MANUAL)
|
|
}
|
|
})
|
|
|
|
// Why an explicit policy rather than nothing: the thread-open path spreads this, so nothing
|
|
// means the fields are ABSENT, and absent is not a reset. A session flipped Yolo → Manual
|
|
// resumed with the Yolo thread's `approvalPolicy: never` still in force and escalated with no
|
|
// prompt. Manual has to say what it wants.
|
|
it('states the approval posture when Manual cleared the flag', () => {
|
|
expect(codexStructuredPermissionPolicyForSettings({ agentDefaultArgs: { codex: '' } })).toEqual(
|
|
MANUAL
|
|
)
|
|
})
|
|
|
|
it('never answers with an absent policy for either posture', () => {
|
|
for (const codex of ['', '--dangerously-bypass-approvals-and-sandbox', '--model gpt-5.6-sol']) {
|
|
expect(
|
|
codexStructuredPermissionPolicyForSettings({ agentDefaultArgs: { codex } }),
|
|
codex
|
|
).toBeDefined()
|
|
}
|
|
})
|
|
|
|
// The passthrough that used to carry these to app-server is gone on purpose; only the
|
|
// permission posture is derived, and nothing else from the field reaches argv.
|
|
it('carries nothing but the permission posture out of the arguments field', () => {
|
|
expect(
|
|
codexStructuredPermissionPolicyForSettings({
|
|
agentDefaultArgs: {
|
|
codex: '--profile review --add-dir /repo -c model_reasoning_effort=high'
|
|
}
|
|
})
|
|
).toEqual(MANUAL)
|
|
})
|
|
})
|