Borehole · Project · free tier
paperclipai/paperclip
Free reading: not for reliance
62c78e5 · 5,551 commits · 11,790 files
Read 2026-10-23

paperclipai/paperclip: technical due diligence, read from its repository

What Borehole read in paperclipai/paperclip's code and history at 62c78e5: its licence, security advisories, releases, review practice, tests, dependencies and operations. Every finding cites the file and line it rests on. The full report also covers the team and asks the questions the code cannot answer.

DimensionRating
Architecture & Codebase — including AI provenancestrong
Engineering Process & SDLC Maturityadequate
Product & Engineering Maturitystrong
Security & Reliability Postureadequate
Observability & Operationsadequate
Fit with the Acquirernot assessable

What the repository says

  • Releases are cut from another branch than the one surveyed
  • AGENTS.md:246: states a policy on AI-generated contributions

Architecture & Codebase — including AI provenance

strong
01.6

server/src/routes/issues.ts is 19,733 lines, 191× this codebase's median file

adequate

The median source file here is 103 lines. 233 files exceed 1,030 lines, the largest being 19,733. Size is not a defect on its own — some problems genuinely live in one place — but these are where merge conflicts, review fatigue and single-owner knowledge concentrate, and they are the first thing that slows an inheriting engineer down.

Evidence
server/src/routes/issues.ts:1 — 19,733 lines (191× median)
server/src/services/native-runtime/native-session-executor.ts:1 — 13,919 lines (135× median)
server/src/services/issues.ts:1 — 13,796 lines (133× median)
server/src/routes/openapi.ts:1 — 12,727 lines (123× median)
server/src/services/tool-gateway.ts:1 — 11,612 lines (112× median)

Engineering Process & SDLC Maturity

adequate
02.16

Tests are retried until they pass

adequate

The suite is configured to retry failing tests (packages/paperclip-runner/src/drivers/codex/codex-question-adapter.test.ts). A retry turns a flaky test green without saying why it failed, so a green build means the tests passed eventually. Ask how many retries a typical run needs, and which tests need them.

02.24

10% of recent commits change more than 2,000 lines

adequate

426 of 4468 commits in the last 24 months changed over 2,000 lines of code, not counting reformatting, renames, vendored or generated code. A change that size is rarely reviewed line by line. Ask how they were reviewed, and what made them that large.

Evidence
426 commits over 2,000 lines

Product & Engineering Maturity

strong
04.62

Releases are cut from another branch than the one surveyed

adequate

The newest release, v2026.1012.0, is not on the branch this survey read; the newest release on it is v2026.1005.0. The shipped product may differ from what this report describes. Survey the release branch too.

Evidence
tag v2026.1012.0

Security & Reliability Posture

adequate
05.6

3 Dockerfiles run the application as root

adequate

No USER instruction switches away from root, so the process runs with full privileges inside the container. It is a one-line fix and an ordinary expectation, which is why its absence is worth asking about: it usually says more about whether anyone has reviewed the deployment than about this container specifically.

05.27

The build runs third-party code it does not pin

adequate

8 third-party CI actions are referenced by a tag or branch rather than a commit, so they run whatever that tag points to on the day; 1 step pipes a download straight into a shell. Tags have been moved to malicious commits in real supply-chain attacks, and a piped script is whatever the server returns that minute. Pinning to a commit is the fix, and it is small.

Evidence
.github/workflows/agent-runtime-images.yml:33 — action pinned by tag
.github/workflows/agent-runtime-images.yml:36 — action pinned by tag
scripts/install.sh:40 — download piped into a shell
05.20

21 SQL statements built from strings

adequate

21 places in product code build SQL by formatting or joining strings rather than passing parameters. In each, what the code shows going in is a table or column name, a constant, or something the line does not show, not a request's value. Ask whether any value reaches these strings, rather than a parameter.

05.26

36 locked dependencies have published vulnerabilities, 27 high or critical in what ships

adequate

36 of 1,978 locked dependencies match 95 published advisories in OSV, 34 of them rated high or critical. A lockfile says which version is installed, not whether the flawed code is reachable from this product, so each needs a look. Most are fixed by an upgrade, and the advisory names the version. This lockfile pins development tools alongside what ships and does not say which is which, so some of these may never reach a user.

Evidence
pnpm-lock.yaml — dompurify 3.4.16: GHSA-cg2g-xh64-v4q7 (critical), fixed in 3.4.17
pnpm-lock.yaml — proxy-addr 2.0.7: GHSA-jqcg-44mw-7w3h (critical), fixed in 2.0.8
packages/plugins/sandbox-providers/daytona/pnpm-lock.yaml — shell-quote 1.10.0: GHSA-pqg4-j6r4-53mv (critical), fixed in 1.11.0
05.9

1 vendored component with no licence

adequate

Third-party code is committed into this repository without any statement of the terms it arrived under. Unknown terms are harder to deal with than inconvenient ones: an awkward licence can be complied with, an absent one cannot. Each of these needs its origin established before the code changes hands.

Evidence
server/src/vendor/paperclip-runner — 3 tracked files, no licence file
05.65

13 advisories against this project in the last 24 months, 10 high or critical

adequate

The newest was published on 2026-08-21. A project that publishes advisories is telling its users about its fixes, and the count and pace say how much security work it carries. Whether 4 of them are fixed in the code you ship could not be told from this history: its version is only known to be at least 2026.1005.0. Answer lives in: the release you run, set against each fixed version. GHSA-265w-rf2w-cjh4 (high), GHSA-3xx2-mqjm-hg9x (critical), GHSA-47wq-cj9q-wpmp (critical) and 6 more are fixed in the surveyed commit; which release carries them cannot be told from this history, which does not place the release users run. Answer lives in: the version you run. The record is incomplete: NVD: 2 records name paperclip with nothing tying them to this repository (CVE-2015-2963, CVE-2017-0889).

Evidence
2026-08-21: CVE-2026-77087 (critical): Paperclip before 0.3.1 in default local_trusted mode fails to validate Host headers, allowing attackers to execute arbitrary commands via DNS rebinding. An atta
2026-07-22: GHSA-x8hx-rhr2-9rf7 (critical): Drive-by RCE Against Local Paperclip Instances via DNS Rebinding
2026-04-16: GHSA-gqqj-85qm-8qhf (high): Paperclip: codex_local inherited ChatGPT/OpenAI-connected Gmail and was able to send real email
2026-04-16: GHSA-265w-rf2w-cjh4 (high), fixed in 2026.416.0: Paperclip: Privilege Escalation via Agent-Controlled workspaceStrategy.provisionCommand Leading to OS Command Execution
2026-04-16: GHSA-3xx2-mqjm-hg9x (critical), fixed in 2026.416.0: Paperclip: Cross-tenant agent API key IDOR in `/agents/:id/keys` routes allows full victim-company compromise

Observability & Operations

adequate
06.35

Sends usage data to telemetry.paperclip.ing by default

adequate

packages/shared/src/telemetry/client.ts sends to telemetry.paperclip.ing, and packages/adapters/codex-local/src/server/parse.ts leaves reporting on as shipped. Every installation reports to the vendor unless its operator turns this off. For a buyer running it inside their own product, that is an outbound data flow to a third party: ask what is sent, how often, and whether the buyer's customers were told.

Evidence
packages/shared/src/telemetry/client.ts:15 — reports to telemetry.paperclip.ing
packages/adapters/codex-local/src/server/parse.ts:69 — the reporting setting is on by default

Fit with the Acquirer

not assessable

Not rated: structurally not assessable from a repository.

Read more, or read your own

The full report on paperclipai/paperclip: every dimension, the questions for the room, and the audit log.

Survey a repository: free on public repositories. Private code is read on your own machine by a collector whose source you can read first.