Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →h5i-dev/h5i
Use when work should run inside a disposable, confined development box instead of on the host — reviewing a pull request or any untrusted or AI-generated code, letting an agent build and test with full autonomy, running a dev server and driving a real browser against it, and exporting the result as a reviewed patch with an execution receipt. Covers creating a box, running commands and interactive sessions in it, reading what a policy actually enforced, and the output gate.
Passed 14 of 20 test cases with the skill, and 4 of 20 without it.
and 11 more in the full report
11 fixed4 passed either way7 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 11 test cases the skill fixed. Same model, same prompt, with and without the skill.
I need to review the complete stream of event logs and execution history recorded for box 'db-migration'. Which command lists the box's event log?
| Checked | Without skill | With skill |
|---|---|---|
| Recommends `h5i box log db-migration` to view the box's event log. | Missed | Passed |
The instructions your model reads when the skill loads.
A box is a disposable development environment: a git worktree on its own branch, confined by a pinned, fail-closed policy. Code, toolchain, dev server and agent run inside it. Nothing reaches the host except what you export.
h5i box <command> --help is the authoritative flag reference and cannot go
stale. Reach for it before guessing at a flag.
$H5I_ENV_ID is set inside one. It changes what you should do:
h5i box # a box from this repository's HEAD
h5i box --pr 1234 # a box from pull request #1234 (number, #n, or URL)
h5i box --name fix-auth # name it yourself; otherwise the branch name is used
Then work with it:
h5i box ls # every box on this clone
h5i box status <name> # policy actually enforced, evidence, base drift
h5i box run <name> -- cargo test # one command, policy-enforced, exit code passes through
h5i box shell <name> # interactive confined session (this is where an agent runs)
h5i box diff <name> # what changed against the pinned base
Every run is recorded as a receipt: the command, its exit code, wall/cpu/
rss, the egress verdicts, and the policy digest that was in force. Secrets are
redacted before anything is written.
test cases passed, vs 4 / 20 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.