## Thinking Path
> - Paperclip manages agent work and its saved files.
> - An ephemeral web server needs remote workspace authority before the
old volume can detach.
> - A task can have a repository request that has never prepared a
checkout.
> - The offline importer currently treats that pending request as
unfinished file publication.
> - This blocks migration even when the target does not exist and there
is no saved commit, branch or failure.
> - This pull request preserves that request in the database and imports
the existing files separately.
> - The next authorized sandbox preparation can acquire the repository
through its existing owner.
## Linked Issues or Issue Description
**What happened?**
On Core `253bf071f0a496554b61fff897ad3f11afaf8766`, offline migration of
a volume-backed workspace fails with `repository_preparation_unresolved`
for an untouched pending repository request. The requested directory has
never existed. The request has no pinned commit, branch or failure.
**Expected behavior**
Preserve the pending request and all existing files. Do not label the
absent checkout as part of a saved archive. Keep partial publication and
active preparation blocked.
**Steps to reproduce**
Create a local task workspace. Add an unmaterialized pending repository
request. Run the canonical offline importer after stopping writers. The
unchanged importer rejects the request before preserving the workspace.
Related migration work: #19429. A GitHub search found no duplicate
pending-repository import fix or matching public issue.
## What Changed
- Admit an untouched pending request only at the existing offline writer
fence and after proving its target is absent.
- Keep the database request unchanged and exclude it from materialized
snapshot descriptors.
- Recheck file absence and source identity before publication. Reject a
request that becomes ready during capture.
- Cover cold filesystem loss, real sandbox-side Git acquisition, and
eleven blocked states. Document the preservation contract.
## Verification
- Red: the new real PostgreSQL/filesystem regression fails on the
unchanged importer with `repository_preparation_unresolved`.
- Green: both affected migration/lifecycle suites pass, **236 tests**.
The cold test deletes the predecessor directory, restores saved bytes,
and acquires the pending repository through the full admission owner,
checking ready state and the pinned commit. Real Git runs against a
local HTTP origin through an exact-URL redirect at the sandbox transport
boundary. It does not claim live Daytona or GitHub qualification.
- Full workspace typecheck and build pass at the final head. Both
affected suites pass **236 tests**, including the full admission owner
and eleven blocked states. All final-head CI checks pass: **54 success,
2 skipped, 1 neutral**. Greptile is **5/5**, with no unresolved review
threads.
- The full local test command failed in `recovers lost inline and check
responses without duplicate GitHub publications` in the chat integration
suite. That exact case passes in a fresh isolated invocation. Its cause
remains unclassified; this is not a full-local-suite pass claim. The
original failure is retained alongside the isolated result.
- No candidate code has been applied to the serving instance.
## Risks
- The exception applies only to offline migration and a pending request
with no target, commit, branch or failure. Unreadable paths, symlinks,
partial directories and active/repair states retain their blockers.
- Requests retain normal repository authorization and acquisition
behavior. This change neither fetches repositories during migration nor
marks requests ready.
- A changed materialized source aborts publication and leaves original
files intact.
## Model Used
OpenAI GPT-6 through Codex, with reasoning and tool use. The exact
deployment variant and context-window size 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
- [x] 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>
