totopo 1.0.5 → 1.0.6
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +16 -15
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -2,7 +2,7 @@
|
|
|
2
2
|
|
|
3
3
|
<img src=".github/assets/logo.png" alt="totopo" width="100%" />
|
|
4
4
|
|
|
5
|
-
A simple CLI to spin up a sandboxed
|
|
5
|
+
A simple CLI to spin up a sandboxed environment for AI coding agents.
|
|
6
6
|
|
|
7
7
|
|
|
8
8
|
|
|
@@ -17,14 +17,6 @@ Reasonable containment for non-deterministic tools. Nothing more, nothing less.
|
|
|
17
17
|
|
|
18
18
|
Note: no sandbox substitutes for good judgment. Consider keeping any sensitive secrets or privileged scripts away from your agents.
|
|
19
19
|
|
|
20
|
-
## Features at a Glance
|
|
21
|
-
|
|
22
|
-
- **Sandboxed Docker container** — your code runs in an isolated environment with strict filesystem and privilege boundaries
|
|
23
|
-
- **Agents can't reach remote** — push, pull, fetch, and clone are blocked inside the container, preventing agents from accidentally affecting your remote repositories
|
|
24
|
-
- **AI CLIs with persistent sessions** — OpenCode, Claude Code and Codex pre-installed, with conversation history that survives restarts and rebuilds
|
|
25
|
-
- **Host-mirror or generic runtime** — use a standard dev container, or let totopo match the container environment to your host so the agent works in the exact same setup as your codebase
|
|
26
|
-
- **Scoped mounts** — expose only the files and directories the agent needs, nothing more
|
|
27
|
-
|
|
28
20
|
## Requirements
|
|
29
21
|
|
|
30
22
|
- [Docker](https://www.docker.com/products/docker-desktop/) - used to build and run the sandboxed environment
|
|
@@ -37,15 +29,24 @@ cd your-project
|
|
|
37
29
|
npx totopo
|
|
38
30
|
```
|
|
39
31
|
|
|
40
|
-
|
|
41
|
-
|
|
32
|
+
First-time setup — running `npx totopo` in a fresh repo, selecting a runtime mode, and waiting for the Docker image to build for the first time:
|
|
42
33
|
|
|
43
|
-
*First-time setup — running `npx totopo` in a fresh repo, selecting a runtime mode, and waiting for the Docker image to build for the first time.*
|
|
44
34
|
|
|
35
|
+
Opening a session when totopo is already initialized is quick. The agent is aware of the sandbox environment:
|
|
45
36
|
|
|
46
|
-
*Starting a session once the container is already built — opening a session, running an AI tool, exiting.*
|
|
47
37
|
|
|
48
|
-
##
|
|
38
|
+
## Core features at a glance
|
|
39
|
+
|
|
40
|
+
- **Sandboxed Docker container** — your code runs in an isolated environment with strict filesystem and privilege boundaries
|
|
41
|
+
- **Agents can't reach remote** — push, pull, fetch, and clone are blocked inside the container, preventing agents from accidentally affecting your remote repositories
|
|
42
|
+
- **AI CLIs with persistent sessions** — OpenCode, Claude Code and Codex pre-installed, with conversation history that survives restarts and rebuilds
|
|
43
|
+
- **Host-mirror or generic runtime** — use a standard dev container, or let totopo match the container environment to your host so the agent works in the exact same setup as your codebase
|
|
44
|
+
- **Agents are sandbox aware** — agents are informed of their sandbox constraints at session start, so they can factor that into how they work.
|
|
45
|
+
- **Scoped mounts** — expose only the files and directories the agent needs
|
|
46
|
+
|
|
47
|
+
---
|
|
48
|
+
|
|
49
|
+
## Features in Detail
|
|
49
50
|
|
|
50
51
|
### Sandboxed dev container
|
|
51
52
|
|
|
@@ -70,8 +71,8 @@ In both scoped modes, `.git` is intentionally not mounted. Mounting `.git` would
|
|
|
70
71
|
|
|
71
72
|
Scoped sessions are well-suited for focused tasks where you want to give the agent a narrow, explicit view of your codebase.
|
|
72
73
|
|
|
74
|
+
Example showcasing agent awareness of scope limitations:
|
|
73
75
|
|
|
74
|
-
*Using scoped mounts — selecting cwd and selective modes, showing what the agent can and can't see inside the container.*
|
|
75
76
|
|
|
76
77
|
### AI CLIs with persistent sessions
|
|
77
78
|
|