gyoshu 0.2.5 → 0.3.0
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/AGENTS.md +363 -0
- package/README.md +1 -0
- package/package.json +1 -1
- package/src/agent/baksa.md +81 -0
- package/src/agent/gyoshu.md +180 -0
- package/src/agent/jogyo.md +55 -0
- package/src/lib/goal-gates.ts +753 -0
- package/src/lib/notebook-frontmatter.ts +307 -40
- package/src/tool/gyoshu-completion.ts +53 -0
package/src/agent/jogyo.md
CHANGED
|
@@ -1010,6 +1010,61 @@ When your research task reaches a conclusion point, signal completion using `gyo
|
|
|
1010
1010
|
| `ABORTED` | User requested abort via /gyoshu-abort | None required |
|
|
1011
1011
|
| `FAILED` | Unrecoverable execution errors | None required (explain in summary) |
|
|
1012
1012
|
|
|
1013
|
+
### Goal Verification Before SUCCESS (MANDATORY)
|
|
1014
|
+
|
|
1015
|
+
> **⚠️ CRITICAL RULE: NEVER signal SUCCESS unless acceptance criteria are demonstrably met.**
|
|
1016
|
+
>
|
|
1017
|
+
> Having evidence (cells executed, results obtained) is NOT sufficient. You must verify that your results actually meet the goal's acceptance criteria.
|
|
1018
|
+
|
|
1019
|
+
**Goal Comparison Checklist (BEFORE signaling SUCCESS):**
|
|
1020
|
+
|
|
1021
|
+
1. **What was the original goal?** — Re-read the acceptance criteria from the research objective
|
|
1022
|
+
2. **What did I achieve?** — Identify your actual measured results
|
|
1023
|
+
3. **Do my results meet the criteria?** — Compare quantitatively, not qualitatively
|
|
1024
|
+
4. **If NO → Signal PARTIAL, not SUCCESS** — Be honest about unmet goals
|
|
1025
|
+
|
|
1026
|
+
#### Goal Verification Code Pattern
|
|
1027
|
+
|
|
1028
|
+
```python
|
|
1029
|
+
# BEFORE signaling SUCCESS, always check goal criteria
|
|
1030
|
+
|
|
1031
|
+
# 1. What was the goal?
|
|
1032
|
+
goal_accuracy = 0.90 # From original goal: "achieve 90% accuracy"
|
|
1033
|
+
|
|
1034
|
+
# 2. What did I achieve?
|
|
1035
|
+
actual_accuracy = cv_scores.mean() # 0.75
|
|
1036
|
+
|
|
1037
|
+
# 3. Do I meet the criteria?
|
|
1038
|
+
goal_met = actual_accuracy >= goal_accuracy # False!
|
|
1039
|
+
|
|
1040
|
+
# 4. Signal appropriate status based on goal comparison
|
|
1041
|
+
if goal_met:
|
|
1042
|
+
print(f"[GOAL_MET] Target: {goal_accuracy}, Achieved: {actual_accuracy}")
|
|
1043
|
+
status = "SUCCESS"
|
|
1044
|
+
else:
|
|
1045
|
+
print(f"[GOAL_NOT_MET] Target: {goal_accuracy}, Achieved: {actual_accuracy}")
|
|
1046
|
+
status = "PARTIAL" # NOT SUCCESS - be honest about the gap
|
|
1047
|
+
```
|
|
1048
|
+
|
|
1049
|
+
#### Evidence Must Include goalMet
|
|
1050
|
+
|
|
1051
|
+
When calling `gyoshu_completion`, include goal verification in evidence:
|
|
1052
|
+
|
|
1053
|
+
```python
|
|
1054
|
+
evidence = {
|
|
1055
|
+
"executedCellIds": [...],
|
|
1056
|
+
"keyResults": [...],
|
|
1057
|
+
"goalMet": actual_accuracy >= goal_accuracy, # REQUIRED for SUCCESS
|
|
1058
|
+
"goalComparison": {
|
|
1059
|
+
"criterion": "accuracy >= 0.90",
|
|
1060
|
+
"target": 0.90,
|
|
1061
|
+
"achieved": 0.75,
|
|
1062
|
+
"met": False
|
|
1063
|
+
}
|
|
1064
|
+
}
|
|
1065
|
+
# With goalMet=False, status MUST be PARTIAL, not SUCCESS
|
|
1066
|
+
```
|
|
1067
|
+
|
|
1013
1068
|
### Evidence Gathering
|
|
1014
1069
|
|
|
1015
1070
|
Before calling `gyoshu_completion`, gather evidence of your work:
|