0 Fork 0

mirror of https://github.com/paperclipai/paperclip.git synced 2026-10-07 07:23:08 +02:00

Files

T

Devin FoleyandPaperclip bac7307ec4 fix(adapter-utils): improve sandbox restore failure diagnostics (#8903 )

## Thinking Path

> - Paperclip is the open source app people use to manage AI agents for
work.
> - Agent adapters share managed-runtime helpers from
`@paperclipai/adapter-utils` so local and sandboxed runs can prepare,
execute, and restore workspaces consistently.
> - The sandbox managed runtime syncs workspaces in both directions by
creating tar archives on the local host and inside the remote sandbox.
> - A prior fix avoided archiving `.` during upload because tar
self-entries can force chmod/utime on a directory the command user does
not own.
> - The restore/download path still archived the remote workspace as
`.`, and failed managed commands only surfaced stderr.
> - This pull request applies entry-based tar creation to the sandbox
restore path and includes stdout in managed command failure diagnostics.
> - The benefit is that restore failures become visible to operators and
sandbox workspace restore avoids the same directory-metadata failure
class already fixed for upload.

## Linked Issues or Issue Description

No existing public issue found; describing in-PR (bug).

- **What happens:** sandbox workspace restore can fail while creating
`workspace-download.tar` from the remote workspace when the tar command
includes a `.` self-entry and the command user cannot update metadata on
the workspace directory. If the failing command writes its diagnostic to
stdout, the managed runtime error can collapse to a generic failed shell
command without the useful tar message.
- **Expected behavior:** restore should archive the workspace entries
without a `.` self-entry, and failed managed runtime commands should
include useful stdout/stderr diagnostics.
- **Where:** `packages/adapter-utils/src/sandbox-managed-runtime.ts`
restore/download path and
`packages/adapter-utils/src/command-managed-runtime.ts` command error
formatting.
- **Related public context:** #7836 fixed the upload side of the same
tar self-entry failure class.

## What Changed

- Added stdout-aware failed-command formatting in the command managed
runtime, keeping diagnostics bounded to the tail of stdout/stderr.
- Added remote workspace tarball creation that names top-level entries
explicitly instead of archiving `.` during sandbox restore.
- Preserved empty-workspace restore support by creating a valid empty
tarball when the remote workspace has no entries.
- Added regression coverage for stdout diagnostics, restore tar members,
and empty workspace restore tarballs.

## Verification

- `pnpm vitest run
packages/adapter-utils/src/command-managed-runtime.test.ts
packages/adapter-utils/src/sandbox-managed-runtime.test.ts` — 2 files,
17 tests passed.
- `pnpm --filter @paperclipai/adapter-utils typecheck` — passed.
- `git diff --check origin/master..HEAD` — passed.
- Local public-hygiene/PII scan of the committed diff checked for
internal ticket refs, local/private URLs, common token/key patterns,
private-key blocks, and email-like values — passed.
- GitHub duplicate/related search found no public issue or PR for
`workspace-download.tar Permission denied` or `sandbox restore tar
permission denied`; #7836 is linked as related prior work.
- PR CI on commit `3ad4c8d` — all GitHub Actions lanes, security scans,
and aggregate `verify` passed.
- Greptile Review on commit `3ad4c8d` — Confidence Score 5/5; the prior
P2 thread is resolved with no open P2s, recommendations, or follow-ups.

## Risks

Low. This is limited to shared adapter runtime error formatting and
sandbox restore archive construction. Archive contents should remain
equivalent apart from the removed `.` self-entry, and the new
diagnostics are bounded to avoid dumping unbounded command output.

## Model Used

OpenAI Codex, GPT-5, tool-enabled coding agent with shell and GitHub CLI
access. Context window size was not reported by the runtime.

## 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 (n/a;
internal adapter-runtime behavior only)
- [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>

2026-07-02 17:06:34 -07:00

src

fix(adapter-utils): improve sandbox restore failure diagnostics (#8903 )

2026-07-02 17:06:34 -07:00

CHANGELOG.md

chore: release v0.3.1

2026-03-12 13:09:22 -05:00

package.json

build(deps-dev): bump @types/node from 22.19.11 to 22.19.21 (#7748 )

2026-06-18 15:46:06 -07:00

README.md

PAPA-430: workspace finalize gates + no-remote-git enforcement (#6969 )

2026-05-29 08:25:29 -07:00

tsconfig.json

Fix root TypeScript solution config

2026-03-09 14:09:30 -05:00

README.md

@paperclipai/adapter-utils

Shared utilities for Paperclip adapters: process spawning, environment injection, sandbox/SSH transport, workspace sync, and the round-trip helpers that move code between the local execution-workspace cwd and wherever the agent actually runs.

For the adapter-author guide see docs/adapters/creating-an-adapter.md and the in-repo notes at packages/adapters/AUTHORING.md.

No-remote-git contract

The local execution-workspace cwd is the only persistence boundary across runs. No adapter may depend on a git remote for cross-run state.

Adapters that run the agent on a different host should use the SSH round-trip helpers in src/ssh.ts:

prepareWorkspaceForSshExecution({ spec, localDir, remoteDir }) — bundles the local cwd (tracked files, dirty edits, untracked additions, and the git history needed to reconstruct it) to remoteDir before the run starts. Runs with no git remote configured.
restoreWorkspaceFromSshExecution({ spec, localDir, remoteDir, ... }) — syncs the remote cwd back into localDir after the run, including any new commits the agent created. Also runs with no git remote configured.

prepareRemoteManagedRuntime in src/remote-managed-runtime.ts wraps both calls for adapters that want a per-run remote workspace and an automatic restoreWorkspace() finally hook.

The invariant is pinned by the no-remote-git contract case in src/ssh-fixture.test.ts, which asserts that a remote-only commit propagates to the local worktree through the prepare → restore round-trip with no git remote configured at any point. Do not regress that test.