DeltaBox Lite

# Fix list: DeltaBox Lite

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/deltabox-lite, the October 2026 research run, assessed 17 October 2026. Grade D, 52.3 out of 100.

This is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public.

For a coding agent working on DeltaBox Lite: work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published.

## 1. Reliability, 37 out of 100, up to 12.6 more on the total

Why it scored 37: Read with the local-software lines, because DeltaBox Lite runs on the owner's machine. Official npm, Homebrew and checksummed release packages for Linux and macOS on x64 and arm64, with Node.js 18, glibc 2.39 and Rust 1.87 stated (20). The repository holds 288 test functions, but its only CI workflow runs on tags or by hand and checks that the packaged binary prints `--version`. No test suite runs in CI, and GitHub's Actions pages were refused to our reader, so no run results were read (5 of 25). Open crash and regression issues could not be read because GitHub's web pages and API answered 403 to our reader, so the line is scored absent (0, unchecked). CHANGELOG.md follows Keep a Changelog and declares semantic versioning, with dated 0.1.0 and 0.1.1 entries. No breaking change has shipped yet, and the 0.1.0 entry still names the old `ddnirvana/tap` Homebrew tap (12 of 15). Version 0.1.1, not declared stable (0). Total 37.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):

Hosted APIs, MCP servers, models and platforms.

- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).
- 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.
- 15, rate limits documented with numbers.
- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.
- 10, an SLA published for any paid tier.
- 10, the surface agents use is generally available, not beta or preview.

Local packages, SDKs, frameworks and stdio MCP servers.

- 20, installs from an official package with supported runtimes stated.
- 25, a public CI and test suite, passing on the default branch.
- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).
- 15, semver discipline and breaking changes called out in a changelog.
- 15, version 1.0 or later, or declared stable.

Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors.

## 2. Security & auth, 47 out of 100, up to 9.3 more on the total

Why it scored 47: No credentials to issue or leak, because the runtime runs on the owner's machine, so the middle band (20 of 30). Least privilege is the product's job. Commands see a confined view, common secret paths are denied, the child environment is scrubbed to nine named variables, writes and pending storage are capped, and changes wait for commit or abort. Against that, the default network mode allows every destination, a `read-write` profile reaches the host with no rollback, macOS `cbcow-through` writes the real project at once, and pi-deltabox documents `isolation: none` as its default (14 of 20). Command output is untrusted content returned to the agent, and we found no guidance for reading it. The agent install prompt itself tells the agent it can star the repository through the user's authenticated GitHub CLI, an action outside the user's request (4 of 15). The session journal records each command's `label` against its change set, and diffs show every changed path. No network log was found for the `allow_audit` mode (9 of 15). No SECURITY.md, security.txt, disclosure policy or published advisories (0 of 20). Total 47.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-security):

- 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.
- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.
- 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.
- 0 to 15, audit logs or per-call visibility for the operator.
- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.

Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing.

## 3. Agent ergonomics, 67 out of 100, up to 5.4 more on the total

Why it scored 67: A CLI adds no tool definitions to an agent's context. `exec` returns stdout, stderr, the exit code, the duration and a `truncated` flag, but each stream may hold up to 8 MiB before truncation (17 of 25). `diff` takes `from` and `to` checkpoints, text diffs stop at 1 MiB per file and conflict reports at 100 paths. There is no paging or path filter on a diff (10 of 20). Every JSON-RPC error uses code -32000 with a message from 15 typed variants, and only a commit conflict carries structured data (12 of 20). Transactions are the product's job. Abort restores the chosen checkpoint, commit is conflict-checked, `session.sync` folds sibling commits, and sessions survive helper restarts. Not an MCP server, so no tool annotations apply (16 of 20). `deltabox run --commit|--abort -- <cmd>` needs no policy or manifest, and the interfaces are a CLI and a Node binding with TypeScript types (12 of 15). Total 67.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):

- 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).
- 20, pagination, filtering and output-size controls.
- 20, actionable, documented error responses, codes and messages an agent can recover from.
- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.
- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.

Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs.

## 4. Transparency & trust, 40 out of 100, up to 5.3 more on the total

Made of editorial 52, provenance 27.

Why it scored 40: Apache-2.0 in LICENSE, Cargo.toml and the npm package (30). The pi-deltabox and dsh-deltabox integration repositories linked from the README were not publicly readable on 17 October, though their npm packages carry Apache-2.0 and MIT. No privacy or data-handling statement. The software is local and the source we read sends nothing to the authors, but nothing says so (10). No deprecation policy, and only two dated changelog entries (2). No telemetry or analytics calls were found in the source, and the docs don't address it (10). Total 52.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):

- 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.
- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).
- 0 to 20, a deprecation policy or notices with dates.
- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).

The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two.

Provenance checks not met in full (half of this category, computed from checked facts):

- Legal entity named: not found (0 of 20)
- Domain age: , no registry record we could read (0 of 15)
- Status page: not found (0 of 10)
- security.txt: not found (0 of 10)

## 5. Schema & documentation, 68 out of 100, up to 5.2 more on the total

Why it scored 68: Read as a CLI and library with a documented stdio protocol. docs/protocol.md lists all 15 JSON-RPC methods with their parameters and result shapes, and the npm package ships TypeScript declarations for the Node binding. There is no JSON Schema or other machine-readable contract for the protocol (15 of 25). No llms.txt. The docs are four Markdown files in the repository, plus a plain-text install prompt written for agents (5 of 10). The README and docs say when to choose each filesystem mode, show the right and wrong forms of shell redirection, and state the path-mapping and macOS limits (15 of 20). `fs_mode`, `durability`, `execution_profile` and review `kind` are closed enums in the protocol and the TypeScript types, and `argv` is an array. `label` is free-form by design (12 of 15). A quick start, a JSON-RPC demo script and a documented `project-conflict` error shape. Other errors share one code and differ only in message text (9 of 15). The protocol is labelled v0.1, with a dated changelog, git tags and GitHub releases (12 of 15). Total 68.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):

APIs and MCP servers.

- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).
- 10, llms.txt or Markdown docs served for agents.
- 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.
- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.
- 0 to 15, examples and documented error responses.
- 15, versioning and a public changelog.

Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference.

## 6. Payments & pricing, 60 out of 100, up to 5 more on the total

Why it scored 60: Read with the self-hosted rule. No x402, MPP or L402 (0 of 40). Apache-2.0 software with nothing to buy and no hosted option, so 20 for pricing, 20 for free use with no card and 20 because an agent can install it from npm or Homebrew with no account. Total 60.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):

The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).

- 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which.
- 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for "contact sales" or prices behind a login.
- 20, a free tier or trial that doesn't need a card.
- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).

Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.

Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol.

## 7. Maintenance & community, 43 out of 100, up to 5 more on the total

Why it scored 43: v0.1.1 on 14 October 2026 (30). Two releases in the last 90 days, v0.1.0 and v0.1.1, both on 14 October, below the three this line needs (0). Issues and pull requests could not be read because GitHub's web pages and API answered 403 to our reader. Git history shows pull request #32 merged on 14 October and commits on 15 October, but reply times were not seen (0, unchecked). A Node-API binding with TypeScript declarations ships in the npm package, in one language (8 of 15). Cargo.lock and a pinned Rust 1.87 toolchain, and a four-platform build on tags. No Dependabot configuration, no test run in CI and no npm provenance attestation (5 of 10). Total 43.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):

- 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.
- 20, at least three releases or dated changelog entries in the last 90 days.
- 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.
- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).
- 10, package health, current dependencies and CI.

Models are read for deprecation notice periods and model churn rather than release counts.

## What we couldn't check

What we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it.

- The scout was right on the product, licence, interfaces, dates and category. One addition is that the repository moved from the Ddnirvana account to the delta-box organisation on 14 October 2026, and the 0.1.0 changelog entry and pi-deltabox's README still name the old account.
- Category code-sandboxes, as suggested. Its main job is a confined place to run an agent's commands with the file changes held for review, the same slot as Microsoft Execution Containers. `sandbox.fs` comes first because the diff, commit and abort view of the filesystem is what distinguishes it.
- install-prompt.txt step 5 asks the installing agent to star the repository with the user's authenticated GitHub CLI. No deduction line covers it, so it is recorded as a security weakness and an agent note. An editor may want to decide whether it counts as a misleading or manipulative claim.
- docs/protocol.md names an MCP server among the adapters, but no MCP server was found in the repository or on npm.
- No legal entity is named. 26 of 105 commits come from sjtu.edu.cn addresses, and the cited paper's authors include several of the committers, but the project does not name an institution.
- unchecked: GitHub's web pages and API answered 403 to our reader, so stars, open issues, reply times, pull requests and Actions run results were not read. Repository facts come from a git clone.
- unchecked: the pi-deltabox and dsh-deltabox GitHub repositories asked for credentials on clone and returned 404 on raw.githubusercontent.com, so their source was not read. Their npm records were.
- unchecked: npm's downloads API returned "package not found" for deltabox-lite on 17 October, likely because the package is three days old, so weekly downloads are not given. www.npmjs.com answered with a bot check and was not read.
- unchecked: nothing was installed or run. Isolation, rollback and the README's latency figures are the project's documentation, not our measurement.

## Weaknesses

- Version 0.1.1, first released on 14 October 2026, with two releases and no stability declaration
- The default network mode `allow_audit` permits unrestricted networking, and no network audit log was found in the source
- The agent install prompt tells the agent it can star the repository through the user's authenticated GitHub CLI
- The only CI workflow builds release binaries and runs `--version` smoke checks on tags. The 288 test functions in the repository are not run in CI
- No SECURITY.md, security.txt or disclosure policy was found

## What costs an agent a turn today

The notes we give agents before they call it. Each one is a workaround an agent shouldn't need.

- Skip step 5 of install-prompt.txt. Starring the repository uses the user's GitHub account and is not part of installing.
- Pass `--commit` or `--abort` to `deltabox run` explicitly. The child's exit status is preserved after that decision.
- Wrap redirection and pipes in `sh -c '...'`. An unquoted `>` is handled by the host shell and writes outside the sandbox.
- Set `[network] mode = "deny_all"` or an `allowlist` in the policy for untrusted code. The default allows every destination.
- Read `resolved_fs_mode`, `fallback_reason` and `degraded_protections` from `session start --json` before relying on isolation.

## What the review panel asked for

- drop the starring step
- run the test suite in CI
- Remove the star step from the install prompt
- Publish a JSON Schema for the protocol
- deny-by-default networking
- remove the starring step

## When it's done

Send what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `"kind": "dispute"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.