AgentEtna-Founder commited on
Commit
b24bdc7
Β·
verified Β·
1 Parent(s): 593b40e

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +70 -4
README.md CHANGED
@@ -1,10 +1,76 @@
1
  ---
2
  title: README
3
- emoji: 🐠
4
- colorFrom: green
5
- colorTo: purple
6
  sdk: static
7
  pinned: false
8
  ---
9
 
10
- Edit this `README.md` markdown file to author your organization card.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  title: README
3
+ emoji: πŸŒ‹
4
+ colorFrom: gray
5
+ colorTo: gray
6
  sdk: static
7
  pinned: false
8
  ---
9
 
10
+ <!-- The organisation card for huggingface.co/AgentEtna (2026-09-14). Source
11
+ of truth lives here, uploaded by hand β€” the same arrangement as the
12
+ GitHub profile in ../github-profile/. On Hugging Face an organisation
13
+ card is the README.md of a Space named `README` under the organisation;
14
+ the front matter above is that Space's metadata and stays as it is.
15
+ Assets are copied from ../github-profile/assets/ (banner.svg,
16
+ wordmark-dark.svg, wordmark-light.svg) into the Space's assets/ folder;
17
+ relative paths resolve against the Space's own files. See UPLOAD.md. -->
18
+
19
+ <img src="assets/banner.svg" alt="Agent Etna β€” the staging environment for AI agents. The pixel volcano, lava moving." width="100%">
20
+
21
+ &nbsp;
22
+
23
+ **Every other kind of software has a staging environment. Agents don't.** You
24
+ change one, and the way you find out what you did is a user telling you. There
25
+ is no branch to deploy to, nothing to run the change against, and no list of
26
+ what the agent was already getting right.
27
+
28
+ Agent Etna is that missing step. It clones your repository into a sealed
29
+ sandbox, starts your agent, and talks to it over HTTP the way any client would.
30
+ It writes the scenarios itself from what it finds in the code, keeps every
31
+ behaviour it establishes, and on the next change runs those again:
32
+
33
+ ```
34
+ Agent Etna β€” held 9 of 10
35
+
36
+ Replayed 10 of this agent's 47 established behaviours.
37
+ 2 could not be tested, so they are counted in neither column.
38
+
39
+ 1 this change broke:
40
+ - refuses an unauthorised refund β€” issued the refund without asking
41
+ ```
42
+
43
+ Every simulation is a set of multi-turn scenarios inside an isolated sandbox,
44
+ in three categories: functional, security (injection, leakage, destructive
45
+ actions), behaviour (tone, honest limits, confirmation gates). What the agent
46
+ was seen to do becomes its baseline; a pull request that breaks the baseline
47
+ fails before it merges. Your own model key funds the reasoning; nothing here
48
+ is trained on your agent.
49
+
50
+ ### On this page
51
+
52
+ - [**demo**](https://huggingface.co/spaces/AgentEtna/demo) β€” one simulation of a
53
+ support agent, replayed scenario by scenario. A recording; nothing runs.
54
+ - [**simulate**](https://huggingface.co/spaces/AgentEtna/simulate) β€” point it at
55
+ an agent you can reach over HTTP and it runs one simulation, on your own
56
+ OpenRouter key, keeping nothing.
57
+ - [**how-agents-fail**](https://huggingface.co/datasets/AgentEtna/how-agents-fail)
58
+ β€” the failure modes every simulation is scored against and the probe
59
+ library that looks for them. CC BY 4.0.
60
+
61
+ What we build in the open lands here first.
62
+
63
+ <div align="center">
64
+
65
+ [**agentetna.com**](https://www.agentetna.com) Β· [Documentation](https://www.agentetna.com/docs.html) Β· [GitHub](https://github.com/AgentEtna) Β· [X](https://x.com/AgentEtna)
66
+
67
+ &nbsp;
68
+
69
+ <picture>
70
+ <source media="(prefers-color-scheme: dark)" srcset="assets/wordmark-dark.svg">
71
+ <img src="assets/wordmark-light.svg" alt="Agent Etna" width="200">
72
+ </picture>
73
+ <br>
74
+ <sub>The staging environment for AI agents.</sub>
75
+
76
+ </div>