Files
PaperClipAI/doc
DottaandPaperclip 60b8089a44 fix(runner): trust managed Cloud sandbox images (#18066)
## Thinking Path

> - Paperclip is the open source app people use to manage AI agents for
work.
> - Cloud selects and updates the sandbox image used by its agents.
> - Runner admission also required saved qualification for the exact
snapshot and configuration.
> - An image update left the old record stale and stopped work before
runtime startup.
> - This pull request trusts the image selected by Cloud and uses the
real runtime startup check.
> - Agents can continue after an image update without an operator
running a separate qualification test.

## Linked Issues or Issue Description

**What happened?**

A managed Cloud image update could cause
`sandbox_snapshot_qualification_required` before provider work began.
Runner discovery also fell back to Legacy when saved platform
qualification was absent. The agent Test action could require a separate
qualification request even when discovery already offered Paperclip
Runner.

**Expected behavior**

A managed Cloud image can start without a saved qualification record.
Runtime startup and the handshake report actual compatibility failures.
Custom images keep their existing qualification rules.

**Steps to reproduce**

1. Use Paperclip Runner on a Paperclip-managed Cloud sandbox.
2. Save qualification for its current snapshot.
3. Update the platform-selected snapshot and leave the old record in
place.
4. Start a task. Before this fix, admission rejects the stale record
before starting the provider.

Related work: Refs #17445, Refs #17724, Refs #17448. These changes cover
saved-computer qualification and setup tests. This fix removes the
qualification prerequisite only for managed Cloud images.

## What Changed

- Use one managed-image policy in runner selection and lease admission.
Require the active Cloud configuration to provision the provider. Active
captured custom images are excluded.
- Use the current Cloud image and ignore missing, stale, revoked or
valid saved qualification records for that image.
- Reuse the existing provisioned-runner selection path. Keep adapter
availability checks and the Legacy choice for managed sandboxes.
- Read custom-image metadata directly during discovery without resolving
provider credentials.
- Use the existing template-matching rule for captured images. Ignore
outdated captures, preserve qualification for matching custom captures,
and fall back to Legacy discovery for malformed saved capture config.
- Skip the separate qualification request when discovery already offers
Paperclip Runner as the default. Test checks the actual runtime.
- Add regressions for image updates, provider failures, custom images,
self-hosted environments and company access. Update the development and
verification guides.

## Verification

- Before the fix, managed-image discovery regressions returned Legacy,
and the rendered Test action called the separate qualification endpoint.
- The original runtime source fails four of the five managed-image
regression cases. The fixed runtime passes all five, including provider
startup error propagation. The baseline check uses the original module
source through a temporary Vitest transform and the same database-backed
tests.
- All 603 focused tests pass: 231 runner selection, 185 environment
runtime, 10 snapshot qualification, 168 rendered agent form and nine
form source controls. Discovery controls cover unavailable credentials,
stale managed markers without an active provisioner, outdated captures,
matching custom captures and malformed saved config. The
outdated-capture regression fails before the fix and passes afterward.
- `node scripts/check-maps.mjs` and `pnpm check:token-gates` pass.
- Full `pnpm -r typecheck` and `pnpm build` pass. All five focused files
passed together on `8275bbe72bb4d91fb3a02fb0658f0846b2abc966`
(`codex/simplify-managed-runner-readiness`, Node 24.19.0, macOS ARM64).
- The full local `pnpm test:run` suite is still in progress. Hosted CI
is waiting at `ci / gate / Select trusted runner`; the separate `review`
workflow is queued. Greptile is 5/5 on the current head, and both review
threads are resolved. This PR is not yet verified by full CI.
- Live fixed-source Daytona acceptance is not verified. This workspace
has no Daytona API key or image configuration. The earlier deployed
recovery is not evidence for this source change.

## Risks

- Managed Cloud now trusts its selected image. A bad image can fail at
real runtime startup instead of a saved qualification gate.
- Company access, credentials, runtime integrity and handshake checks
still apply. Self-hosted sandboxes, custom environments and active
captured custom images keep their qualification and immutable snapshot
binding.
- No schema change or data migration is required. Existing qualification
records can remain in place.

## Model Used

- OpenAI Codex, GPT-6 family, with reasoning, tool use and code
execution. The exact model ID and context window are not exposed in this
session.

## Checklist

- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [ ] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge

---------

Co-authored-by: Paperclip <noreply@paperclip.ing>
2026-10-28 14:39:26 -05:00
..
2026-02-16 14:25:00 -06:00