cc-safety-net

A pre-execution guard for AI coding agents. It blocks destructive Git and file system commands, plus common attempts to access sensitive files, before a tool call runs. Supports Amp Code, Antigravity CLI, Claude Code, Codex, Cursor, DeepSeek Harness, Devin CLI, GitHub Copilot CLI, Grok Build, Hermes Agent, Kimi Code, OpenClaw, OpenCode, and Pi.

Runssource: GitHubhomepageTypeScriptMITcommit af34a34ba0a0

TypeScript, MIT licensed. The project labels itself: ai agents, ai safety, antigravity, claude, claude code, claude code plugin, codex and codex cli.

cc-safety-net runs. An Argusic agent installed it in 2 minutes and hit 5 errors and fixed every one without help on a clean machine with no GPU, and the whole session was recorded.

The cc-safety-net project installs, builds, and runs its full test suite (3674 passing, 1 pre-existing failure). The built CLI binary answers version 2.4.6, the hook blocks destructive commands like 'git reset --hard', the explain command displays blocked verdicts, status and doctor produce reports, the API entry returns deny/allow decisions, and the OpenCode, Pi, and Hermes Agent e2e integration tests pass when buil...

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 93.3 of 100 (the mean of this project's run scores).

At a glance

verdict
Runs
Argusic Score
93.3 / 100
cost of the verifying run
$1.31 (measured)
recorded runs
2
last tested
stars
2,081
forks
98
open issues
3
watchers
9
size
57 MB
created
last push

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.

Time lost to each failure while testing cc-safety-net, drawn from the recorded run
Every bar is a failure the agent recorded, with the time it cost. Drawn from this run, not from anywhere else.

What went wrong, in the agent's own words

The agent's own words, worst first, never rewritten into advice.

  • 10 test failures (6 packed-runtime, 2 Pi e2e, 2 OpenCode e2e, 2 Hermes e2e) caused by buildRuntimeBundles not writing a package.json with type:module at the output root — Node.js could not load ES module output files in temp directories15 minutes
  • linked worktree facts test failure — git config parser didn't detect unterminated double-quoted strings (backslash-escaped final quote), returning a wrong path instead of null like real git5 minutes
  • atomic policy writes test failure — writeFileSync respects umask (002), creating the test fixture at 0o664 instead of expected 0o6443 minutes
  • Bun was not pre-installed in the environment; needed unzip which wasn't available apt-get installable without root2 minutes
  • text doctor shows active custom rules — pre-existing failure (exit code 1 instead of 0) not caused by my changes1 minute
Install time
2 minutes
Cold machine to finish
34 minutes
Errors hit and fixed
5 hit, 5 fixed with no human help
How the result was proved
bun test ran 3674 pass, 2 skip, 1 fail (pre-existing), 49 snapshots, 266461 expect() calls across 215 files; node dist/bin/cc-safety-net.js --version printed 2.4.6; bun run check passed lint, format, typecheck, knip, duplicates, coverage, and verify
Model tokens used
438,365
Exact commit tested
af34a34ba0a0
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

The end of the recording

The last lines of the terminal session that decided this verdict, exactly as they were recorded.

(pass) status renders a report [171.46ms]
(pass) doctor renders a parseable report [222.49ms]
(pass) the api entry answers a deny and an allow [152.71ms]
(pass) the index entry exports the OpenCode plugin [184.52ms]
 8 pass
 0 fail
 13 expect() calls
Ran 8 tests across 1 file. [1.54s]
Now I have all the information I need for the report.
tokens used
438,365
Now I have all the information I need for the report.

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.
  • Installed in 2 minutes, faster than the median of the 123 comparable projects Argusic has measured.
  • Recovered from all 5 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.
  • Licensed MIT, as reported by its host.
  • Measured 2 times, so the result is not a one-off.

Of the 124 ai-agents projects Argusic has installed and timed, cc-safety-net was the 35th fastest to reach a running state, and 96 of 124 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.
  • A quick evaluation: it was running 2 minutes after the clone.

Also tested, in the same area

Every one of these was installed and run by Argusic on a clean machine. Nothing appears here that was not tested.

Run history

tested byattemptoutcomescorecost USDwhen
Argusic Runner2/3Runs100.001.31
Argusic Runner1/3Runs86.670.19

Topics (from GitHub)

ai-agentsai-safetyantigravityclaudeclaude-codeclaude-code-plugincodexcodex-clicoding-agentcursordeveloper-toolsdshdsh-pluginguardrailshermes-agenthooksopenclawopencodepi-extensionsecurity

Embed the badge

Markdown for the project README. It links back here; terms on the terms page.

[![Tested by Argusic](https://argusic.com/badge/cc-safety-net.svg)](https://argusic.com/subject/cc-safety-net)

Questions

Does cc-safety-net run?
Yes. cc-safety-net runs. Argusic installed and launched it on a clean machine in 2 minutes, hitting 5 errors on the way, and recorded the session.
How did Argusic test cc-safety-net?
On a fresh, disposable machine, with every command recorded and the repository pinned at commit af34a34ba0a0. 2 attempts are recorded, and the full method is on the methodology page.
What did it cost to test cc-safety-net?
The run that produced this verdict cost $1.31: real compute and model cost, not a list price per million tokens. It is recorded on that run's page.
How long does cc-safety-net take to install?
2 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 cc-safety-net need a GPU?
Not to start. Argusic reached a running state on a machine with no GPU.
What goes wrong when installing cc-safety-net?
5 things broke in the recorded run, and 5 were fixed without human help. Each one, and the time it cost, is listed on this page.
How does cc-safety-net compare with the alternatives?
Of the 124 ai-agents projects Argusic has installed and timed, cc-safety-net was the 35th fastest to reach a running state, and 96 of 124 reached one at all.
Where is the evidence for cc-safety-net?
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.

Discussion