x_aeon_agents 2.6.0 → 2.6.1
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.
- checksums.yaml +4 -4
- data/CHANGELOG.md +6 -0
- data/TODO.md +0 -7
- data/lib/x_aeon_agents/agents/plan_generator_agent.rb +1 -0
- data/lib/x_aeon_agents/agents/tester_agent.rb +22 -17
- data/lib/x_aeon_agents/version.rb +1 -1
- metadata +1 -1
checksums.yaml
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
---
|
|
2
2
|
SHA256:
|
|
3
|
-
metadata.gz:
|
|
4
|
-
data.tar.gz:
|
|
3
|
+
metadata.gz: 9e35c93ce34d5bcbd4425bd104b6f7e318a5d80d32176d9c4292d886acf25ae5
|
|
4
|
+
data.tar.gz: 425cc6c7babc621315dddba419cf3a0e331104d902b4794ea7bab3ea979231fe
|
|
5
5
|
SHA512:
|
|
6
|
-
metadata.gz:
|
|
7
|
-
data.tar.gz:
|
|
6
|
+
metadata.gz: a0eca635a5b4a026f14c88d1bc041fe24d78355fa06c2a0dd00644df0a7c2d6265a7c20ecfed8ffd16c8075e4e9a84e9486b3a2e1dcb9abf0e0ddc8eee57c40d
|
|
7
|
+
data.tar.gz: c886025ea23528c114403da962f878cf8d2176a778c81211fdac2fe98881c44ea29310c588c3ec758199b1cbc9357822cefc2beb769c538cf4e1c601eaa419fb
|
data/CHANGELOG.md
CHANGED
|
@@ -1,3 +1,9 @@
|
|
|
1
|
+
# [v2.6.1](https://github.com/Muriel-Salvan/x_aeon_agents/compare/v2.6.0...v2.6.1) (2026-11-07 01:29:31)
|
|
2
|
+
|
|
3
|
+
### Patches
|
|
4
|
+
|
|
5
|
+
* [refactor(agents): improve prompts to avoid artifact files](https://github.com/Muriel-Salvan/x_aeon_agents/commit/71afa833c80e574eb74b6349e6371762a447350a)
|
|
6
|
+
|
|
1
7
|
# [v2.6.0](https://github.com/Muriel-Salvan/x_aeon_agents/compare/v2.5.3...v2.6.0) (2026-11-06 19:39:17)
|
|
2
8
|
|
|
3
9
|
## Global changes
|
data/TODO.md
CHANGED
|
@@ -1,10 +1,3 @@
|
|
|
1
|
-
* Improve prompts:
|
|
2
|
-
- "Never create files (`ARTIFACT_*`, `/tmp/*`) to hold artifacts. Put the fenced ` ```json output_artifact=… ` block in your final message and end the turn with __no tool call__. If `ARTIFACT_PLAN_MODIFICATIONS` has nothing to report, say so in plain text."
|
|
3
|
-
* Improve Tester prompt:
|
|
4
|
-
- "Fix __all__ issues reported in `ARTIFACT_TESTS_OUTPUT` in this iteration (not a subset)."
|
|
5
|
-
- "Identify the exact failing spec file(s) and run __those__ (`<tests_cmd> spec/scenarios/code_quality_spec.rb`), then a final full run. __Never__ conclude success from a partial run."
|
|
6
|
-
- "Quote the final summary line of your verification run in your answer."
|
|
7
|
-
* Replace `JSON.dump(artifact_content)` by `JSON.pretty_generate(artifact_content)` (and `JSON.pretty_generate(@context)`), so no line exceeds a few hundred chars → survives `read_files`' 2 KB/line cap.
|
|
8
1
|
* Make the tester loop with a max retries when the same tests are failing (maybe use a JSON/XML output format for tests that allow for parsing which test fail and only give this information in the prompt).
|
|
9
2
|
* Don't use project skills and rules: it adds noise to the LLMs that scan the files and code base. Use only global objects from dedicated config dirs.
|
|
10
3
|
* Add desktop notification when user input is needed.
|
|
@@ -47,6 +47,7 @@ module XAeonAgents
|
|
|
47
47
|
- You may only analyze and propose plans.
|
|
48
48
|
- Do NOT execute the plan yourself.
|
|
49
49
|
- #{Helpers.repository_access_constraint}
|
|
50
|
+
- Never create files (`ARTIFACT_*`, `/tmp/*`) to hold artifacts. Put the fenced ` ```json output_artifact=… ` block in your final message and end the turn with __no tool call__.
|
|
50
51
|
EO_CONSTRAINTS
|
|
51
52
|
**agent_params
|
|
52
53
|
)
|
|
@@ -42,28 +42,33 @@ module XAeonAgents
|
|
|
42
42
|
Fix any regression that has been induced by new features or fixes, while keeping the initial requirements and implementation plan in mind.
|
|
43
43
|
If the decisions taken in the implementation plan prevent you from fixing regressions, modify the implementation plan and report those modifications to the user.
|
|
44
44
|
EO_OBJECTIVE
|
|
45
|
-
system_instructions: <<~EO_INSTRUCTIONS,
|
|
46
|
-
#{Helpers.repository_access_system_instructions}
|
|
47
|
-
|
|
48
|
-
Always gather all the necessary context before starting to work on a task.
|
|
49
|
-
For example, make sure you understand the regressions to fix, the naming conventions, frameworks and libraries used and aligned in the current codebase, and the environment and commands used to run and test the code etc.
|
|
50
|
-
Always run the whole tests suite using the tests command provided, and validate your fixes with live feedback until all tests pass.
|
|
51
|
-
Begin by analyzing the user's input and gathering any necessary additional context.
|
|
52
|
-
Then, present your plan at the start of your response along with tool calls before proceeding with the task.
|
|
53
|
-
It's OK for this section to be quite long.
|
|
54
|
-
|
|
55
|
-
Review each question carefully and answer it with detailed, accurate information.
|
|
56
|
-
|
|
57
|
-
If you need more information, use one of the available tools or ask for clarification instead of making assumptions or lies.
|
|
58
|
-
|
|
59
|
-
When you have completed the task, please provide a summary of what you did and any relevant information that the user should know, including the final summary line of the tests suite.
|
|
60
|
-
This will help ensure that the user understands the changes made and can easily follow up if they have any questions or need further assistance.
|
|
61
|
-
EO_INSTRUCTIONS
|
|
62
45
|
constraints: <<~EO_CONSTRAINTS,
|
|
63
46
|
- #{Helpers.repository_access_constraint(writable: true)}
|
|
47
|
+
- Never create files (`ARTIFACT_*`, `/tmp/*`) to hold artifacts. Put the fenced ` ```json output_artifact=… ` block in your final message and end the turn with __no tool call__.
|
|
64
48
|
EO_CONSTRAINTS
|
|
65
49
|
**agent_params
|
|
66
50
|
)
|
|
51
|
+
self.system_instructions = <<~EO_INSTRUCTIONS
|
|
52
|
+
#{Helpers.repository_access_system_instructions}
|
|
53
|
+
|
|
54
|
+
Always gather all the necessary context before starting to work on a task.
|
|
55
|
+
For example, make sure you understand the regressions to fix, the naming conventions, frameworks and libraries used and aligned in the current codebase, and the environment and commands used to run and test the code etc.
|
|
56
|
+
Always run the whole tests suite using the tests command provided, and validate your fixes with live feedback until all tests pass.
|
|
57
|
+
Begin by analyzing the user's input and gathering any necessary additional context.
|
|
58
|
+
Then, present your plan at the start of your response along with tool calls before proceeding with the task.
|
|
59
|
+
It's OK for this section to be quite long.
|
|
60
|
+
|
|
61
|
+
Review each question carefully and answer it with detailed, accurate information.
|
|
62
|
+
|
|
63
|
+
If you need more information, use one of the available tools or ask for clarification instead of making assumptions or lies.
|
|
64
|
+
|
|
65
|
+
Fix __all__ issues reported in the `#{artifact_ref(:tests_output)}` artifact in this iteration (not a subset).
|
|
66
|
+
Identify the exact failing spec file(s) and run __those__, then a final full run. __Never__ conclude success from a partial run.
|
|
67
|
+
Quote the final summary line of your verification run in your answer.
|
|
68
|
+
|
|
69
|
+
When you have completed the task, please provide a summary of what you did and any relevant information that the user should know, including the final summary line of the tests suite.
|
|
70
|
+
This will help ensure that the user understands the changes made and can easily follow up if they have any questions or need further assistance.
|
|
71
|
+
EO_INSTRUCTIONS
|
|
67
72
|
end
|
|
68
73
|
end
|
|
69
74
|
end
|