prime-agent
A self-improving RLM agent for coding workflows and long-running autonomous tasks.
Runssource: GitHubRustNOASSERTIONcommit e260085dd8f7
Rust, NOASSERTION licensed.
prime-agent runs. An Argusic agent installed it in 18 minutes and hit 7 errors and fixed every one without help on a clean machine with no GPU, and the whole session was recorded.
With Node 22.17.0, uv 0.13.6, lsof 4.99.7, and real catalog assets in place, the monorepo installs, builds all four packages, launches the prime-agent CLI (--version 0.9.6), boots its Python kernel venv, and passes 4843+509+66+538 tests with 0 failures, after two source fixes (extension loader dist fallback and lsof socket-path suffix stripping) and updated catalog/data-dependent test expectations.
What the agent ended up with on a clean machine, in its own words. How this is measured
Measured by Argusic on a fresh machine every time. Every number links to its evidence. Argusic Score 96 of 100 (the mean of this project's run scores).
At a glance
Subject data from GitHub, linked at the top of this page, refreshed . Test data by Argusic (CC BY 4.0); every number links to a run page with the full log, the recording, and their sha256 hashes.
What was measured
From the run that decided this verdict. Every line is a recorded value, not an estimate.
What went wrong, in the agent's own words
The agent's own words, worst first, never rewritten into advice.
- lsof missing on host (container has no ss either), breaking a shutdown --force regression30 minutes
- Extension loader aliased pi packages to unbuilt dist/ paths, breaking extension import and 10 tests25 minutes
- 9 tests pinned stale catalog data or wrong no-hang assertions (catalog grew 68->69, gpt-5.4 became image-capable, bundled models polluted hermetic harness tests, stdin guard now exits 0)25 minutes
- uv missing, required for the Python kernel bootstrap6 minutes
- No prime-agent-catalog checkout; tests need generated bundled assets6 minutes
- npm run check fails at check:test-policy with pre-existing prime-agent-runtime Python test violations5 minutes
- Install time
- 18 minutes
- Cold machine to finish
- 65 minutes
- Errors hit and fixed
- 7 hit, 7 fixed with no human help
- How the result was proved
- cd packages/coding-agent && npm test exited 0 with 4843 passed, 52 skipped, 0 failed across 282 files; packages/ai vitest exited 0 (509 passed, 196 skipped); packages/agent npm test exited 0 (66 passed); packages/tui npm test exited 0 (538 passed); ./prime-agent.sh --version printed 0.9.6; ./prime-agent.sh shutdown --force printed 'stopped background service (pid 4762)'
- Model tokens used
- 3,574,399
- Exact commit tested
- e260085dd8f7
- Container image
- sha256:cdd920bce783
What was not measured
The other half, so you can see exactly how far this evidence goes. A gap is printed as a gap and never filled in.
- Peak GPU memory
- this ran on a CPU machine, so there was no GPU to sample
- Behaviour over a long run
- each attempt uses the smallest documented settings and a fixed time budget, so this measures whether it runs, not how it holds up under load
Watch the session
The whole run, from a clean machine, at the speed it actually happened. Nothing is cut and nothing is sped up.
A recording of attempt 2, run , the attempt that decided the verdict above. The terminal recording it was made from is the evidence, stored with a sha256 fingerprint; the video is a rendering of it. Nothing loads from the video host until you press play.
The end of the recording
The last lines of the terminal session that decided this verdict, exactly as they were recorded.
- `packages/coding-agent/src/cli/daemon-ps.ts` — Linux `lsof -F n` emits a ` type=STREAM` suffix; `… Test expectations updated to current catalog/data (all verified against the regenerated real catalo… - `test/mcp-service-catalog.test.ts` — shipped catalog pin 68 → 69 (current real catalog). - `test/image-model-override.test.ts` — "unusable imageModel" now uses an id absent from the catalo… - `test/stdin-guard-cold-cli.test.ts` — the fixed guard gives up on the silent stdin pipe (message … - `test/suite/agent-session-runtime.test.ts` and `test/suite/regressions/4649-subagent-model-select… Verification results: - `npm ci` succeeded (432 packages). - Builds: tui, ai, agent, coding-agent all built; `dist/bundle/cli.js` produced. - `--version` → `0.9.6`; TUI launched on the virtual display showing the banner, model, cwd, and in… - Suites: ai 509 passed, agent 66 passed, tui 538 passed, coding-agent 4843 passed (282 files), 0 f… - `npm run check` fails only at `check:test-policy` with pre-existing violations in `prime-agent-ru…
Replay the whole session, every command from a clean machine to this point.
How it was tested
One agent, one container, no human help. It installs the project the way its own documentation says to, runs it, and fixes what breaks. Everything below is recorded as it happened: the terminal session, the log and the exact commit. The full method.
Strengths and limits
Measured facts, not opinions. How this is written.
What went well
- Reached a running state on a clean machine, with the session recorded.
- Recovered from all 7 errors without a human stepping in, which says the failures are documented well enough to solve.
- Ran without a GPU, so it does not need one to start.
- Measured 2 times, so the result is not a one-off.
What did not, or is not known
- Took 18 minutes to install, slower than the median of the 34 comparable projects Argusic has measured.
- Its license file could not be classified automatically, so check it yourself before using it.
Of the 35 TypeScript projects Argusic has installed and timed, prime-agent was the 32nd fastest to reach a running state, and 28 of 35 reached one at all.
What it is a reasonable choice for
- Trying it on a laptop or a small server: it reached a running state without a GPU.
Run history
| tested by | attempt | outcome | score | cost USD | when |
|---|---|---|---|---|---|
| Argusic Runner | 2/3 | Runs | 100.00 | 0.87 | |
| Argusic Runner | 1/3 | Runs with mocks | 92.00 | 0.81 |
Embed the badge
Markdown for the project README. It links back here; terms on the terms page.
[](https://argusic.com/subject/prime-agent)Questions
- Does prime-agent run?
- Yes. prime-agent runs. Argusic installed and launched it on a clean machine in 18 minutes, hitting 7 errors on the way, and recorded the session.
- How did Argusic test prime-agent?
- On a fresh, disposable machine, with every command recorded and the repository pinned at commit e260085dd8f7. 2 attempts are recorded, and the full method is on the methodology page.
- What did it cost to test prime-agent?
- The run that produced this verdict cost $0.87: real compute and model cost, not a list price per million tokens. It is recorded on that run's page.
- How long does prime-agent take to install?
- 18 minutes in the run that decided this verdict, from a clean machine to a running state. That is one measured attempt, not an average.
- Does prime-agent need a GPU?
- Not to start. Argusic reached a running state on a machine with no GPU.
- What goes wrong when installing prime-agent?
- 7 things broke in the recorded run, and 7 were fixed without human help. Each one, and the time it cost, is listed on this page.
- How does prime-agent compare with the alternatives?
- Of the 35 TypeScript projects Argusic has installed and timed, prime-agent was the 32nd fastest to reach a running state, and 28 of 35 reached one at all.
- Where is the evidence for prime-agent?
- All 2 recorded runs are on this page, each linking to its full log and terminal recording, stored with a sha256 fingerprint so it cannot be quietly altered.