Prime Agent

# Fix list: Prime Agent

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/prime-agent, the October 2026 research run, assessed 8 October 2026. Grade C, 60.5 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 Prime Agent: 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. Schema & documentation, 42 out of 100, up to 9.4 more on the total

Why it scored 42: Harness reading of the framework line. `--mode json` emits a session event stream and `rpc` and `acp` modes exist, but we found no published schema or reference for them outside the source (10). docs.primeintellect.ai serves an llms.txt that has no Prime Agent page (0). The README states what the agent does, how it differs and that it is not a sandbox, and the bundled skills say when to use each function (10). CLI values are validated with messages that list the valid choices, and MCP tool schemas are read live from the server (7). The README gives the install line and seven commands, and the MCP skill documents its error types, but there is no headless or configuration guide (5). Versioned releases with notes for stable versions, none of substance for nightlies (10).

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.

## 2. Security & auth, 50 out of 100, up to 8.8 more on the total

Why it scored 50: Harness reading of the framework checklist, as used for the other harnesses. 30 for what leaves the machine by default, 20 for approvals and sandboxing, 15 for prompt-injection posture, 15 for audit and 20 for the security programme. Usage telemetry is on by default. It is pseudonymous, limited to primitive values with no prompts, tool content or file paths per the source, disclosed on first run and mirrored to a local `telemetry.jsonl` (18). No approval prompts were found and there is no sandbox, which the README states. MCP tools can be limited with `enabledTools` and `disabledTools`, and the agent is told to recommend a connection, not make one (4). The README tells users to work only with trusted repositories, instructions and skills. No detection or mitigation found (5). Append-only JSONL session files, a `prime-agent incident` timeline command and the local telemetry mirror (12). The vendor's security policy has a reporting address, safe harbour, response times and discretionary rewards, scoped to its website and API. The repository has no SECURITY.md and no published advisories, CONTRIBUTING.md says to report privately without naming an address, and security.txt returned 404. cargo-deny runs in CI (11).

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. Reliability, 65 out of 100, up to 7 more on the total

Why it scored 65: Read with the local-software lines. The vendor's installer fetches checksummed binaries for macOS and Linux on x64 and arm64, with an npm install needing Node.js 20.6 or later as the fallback, but the stable pointer is the TypeScript build while the Rust code on main ships only on the nightly channel, and Windows builds are beta only (17). A public CI workflow runs fmt, clippy with warnings as errors, sharded workspace tests and cargo-deny, and its badge read passing on 8 October 2026 (25). 20 open issues and 91 open pull requests on 8 October. Issues hold only work the maintainers accepted, and the Discussions bug category lists reports from the past two weeks of session-worker crashes, truncated RPC output and provider hangs, whose reply state we did not read (15). Stable releases carry written notes and changes are recorded as fragments in `.changes`, including the removal of the `--tools` flags, but there is no changelog file, the nightly notes are boilerplate and the project is at 0.x (8). Version 0.9.8, and the Rust build is a beta (0).

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.

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

Why it scored 60: Self-hosted rule. No payment protocol (0). Free and MIT-licensed with nothing to buy, so 20, 20 and 20 on the last three lines. The model provider is paid separately.

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.

## 5. Agent ergonomics, 73 out of 100, up to 4.4 more on the total

Why it scored 73: Harness reading, scored on what a person or pipeline driving it has to supply. The model sees three tools (bash, edit, ipython), and MCP tools are searched, described and called from Python instead of being loaded into the prompt (23). Autonomous mode has limits for turns, tokens, time and continuations, a goal token budget and automatic compaction (17). A JSON event stream and exit codes for headless runs, described in the crate README and source only (10). Sessions resume, the daemon restarts workers and restores from JSONL files, refinements can be rolled back and quality gates retry (15). One command starts it and no SDK was found. A provider must be chosen through an interactive `/login` or supplied as a key first (8).

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.

## 6. Transparency & trust, 64 out of 100, up to 3.2 more on the total

Made of editorial 67, provenance 61.

Why it scored 64: MIT in the LICENSE file and Cargo.toml, with a notice that the code is a port of a TypeScript product originally copyright Mario Zechner (30). The first-run notice and the telemetry crate README say what is sent and what is excluded, and name the endpoint and PostHog, but the vendor privacy policy is dated 23 February 2024, does not describe agent telemetry and gives no retention period for it (12). No deprecation policy found. Removed flags are recorded in change fragments (5). Telemetry is disclosed with a setting, a command and two environment variables to turn it off (20).

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):

- Domain age: primeintellect.ai, no registry record we could read (0 of 15)
- Terms of service: read, states 5 of the 7 things a reader expects, and has 3 clauses that cost points (2.3 of 10)
- Privacy policy: read, states 7 of the 8 things a reader expects (9.3 of 10)
- security.txt: not found (0 of 10)

## 7. Maintenance & community, 79 out of 100, up to 1.8 more on the total

Why it scored 79: Nightly v0.9.9-beta.55 on 7 October 2026 and stable v0.9.8 on 29 September 2026 (30). At least 12 stable tags since 20 August 2026 and 55 nightly betas since 30 September (20). The main branch had at least 200 commits between 22 September and 8 October, and Discussions had new threads daily in October, but issues from unapproved contributors and unsolicited pull requests are closed automatically and we could not read reply times (14). Binaries for five targets from one release workflow and Homebrew definitions in the repository. No SDK and no registry package found (8). A pinned toolchain, a committed lockfile, cargo-deny with two triaged advisory exceptions reviewed on 17 September 2026, and CI passing. An open issue (#3292) reports npm audit findings for the 0.9.8 install, which we did not verify (7).

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.

- Unchecked: reply times and maintainer answers in Discussions and on the 20 open issues (the GitHub API was rate-limited for our address on 8 October, and the HTML lists do not show replies)
- Unchecked: the full release list beyond the first page, so the count of stable releases in 90 days comes from git tags in a shallow clone (12 from v0.7.4 on 20 August 2026)
- Unchecked: whether trace sharing through `/traces` is off until the user turns it on. The source describes a settings flag and we did not trace its default
- Unchecked: the npm audit findings reported in issue #3292 for the 0.9.8 install
- Unchecked: whether the Homebrew cask and formula in the repository are published to a tap
- The scout wrote that every GitHub release is a pre-release. v0.9.8 of 29 September 2026 is marked Latest and is a stable release, of the TypeScript build. The Rust build on main is nightly only
- The scout wrote that the README does not mention MCP or a headless mode. That is right for the README, but the source has an MCP client, `-p` print mode and `--mode json|rpc|acp`, none documented for users
- The LICENSE names Prime Intellect Ltd. while the website and terms name Prime Intellect, Inc. Which entity publishes the agent is not stated
- No documentation was found for the JSON event stream, the RPC protocol or the settings file

## Weaknesses

- No approval prompts and no sandbox. The README says the worker and kernel processes are not a security sandbox
- The stable channel (v0.9.8, 29 September 2026) is the TypeScript build. The Rust code on main ships only as nightly betas
- No documentation site for Prime Agent. Headless, JSON, RPC and ACP modes are described only in source and crate READMEs
- Usage telemetry is on by default, and the vendor privacy policy (updated 23 February 2024) does not describe it
- No SECURITY.md in the repository, and public issues and unsolicited pull requests are closed automatically

## 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.

- Run it in a disposable clone, container or VM. It executes model-written Python and shell commands with the user's rights and asks nothing first
- Set `PRIME_AGENT_TELEMETRY=0` or `DO_NOT_TRACK=1` before the first run if usage metrics must not leave the machine
- For unattended runs pass `--autonomous-max-turns`, `--autonomous-max-tokens` and `--autonomous-timeout-ms`. Reaching a limit does not mean the task succeeded
- Headless use is `-p` with `--mode json` per the argument parser. A provider key must already be configured, because `/login` is interactive
- The `-t/--tools`, `-nt/--no-tools` and `-nbt/--no-builtin-tools` flags were removed and now fail as unknown options

## 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.