kairn-cli 2.2.7 → 2.2.8
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/dist/cli.js +7 -2
- package/dist/cli.js.map +1 -1
- package/package.json +1 -1
package/dist/cli.js
CHANGED
|
@@ -4748,6 +4748,13 @@ the agent lacks a tool it needs, or has tools that add noise without benefit.
|
|
|
4748
4748
|
- Consider both additions AND removals. Remove sections that add noise without improving task performance.
|
|
4749
4749
|
- Bloated harnesses hurt performance \u2014 trim what isn't earning its keep.
|
|
4750
4750
|
|
|
4751
|
+
## Anti-Gaming (CRITICAL)
|
|
4752
|
+
- Mutations must improve GENERAL-PURPOSE development quality, not target specific eval criteria.
|
|
4753
|
+
- You do NOT have access to scoring rubrics or expected outcomes. Diagnose problems from traces only.
|
|
4754
|
+
- Do NOT add over-specified rules that restate existing conventions with stronger emphasis (e.g., changing "use chalk.green for success" to "MUST use chalk.green, no exceptions"). If a convention already exists, trust it.
|
|
4755
|
+
- Do NOT add rules that only apply to a narrow eval scenario (e.g., write permissions for a specific directory just because one task needed it).
|
|
4756
|
+
- Ask: "Would this mutation help a developer working on ANY task in this project?" If not, don't propose it.
|
|
4757
|
+
|
|
4751
4758
|
Return ONLY valid JSON.`;
|
|
4752
4759
|
var STDOUT_TRUNCATION_LIMIT = 1e3;
|
|
4753
4760
|
var MAX_CONTEXT_CHARS = 1e5;
|
|
@@ -4806,8 +4813,6 @@ ${content}
|
|
|
4806
4813
|
`### Task: ${task.id}
|
|
4807
4814
|
- Template: ${task.template}
|
|
4808
4815
|
- Description: ${task.description}
|
|
4809
|
-
- Expected outcome: ${Array.isArray(task.expected_outcome) ? task.expected_outcome.join("; ") : task.expected_outcome}
|
|
4810
|
-
- Scoring: ${task.scoring}
|
|
4811
4816
|
`
|
|
4812
4817
|
);
|
|
4813
4818
|
}
|