learn-timeout
General-purpose timeout estimation skill. Trains dual models (duration regression + risk classification) from corpus data and observation feedback. Returns calibrated timeout predictions with confidence intervals for any task type.
소스 정보
- 저장소
- grahama1970/agent-stack-public
- 최근 소스 활동
- 2026년 9월 24일 15:51
- 감지된 SKILL.md 언어
- 영어
- 스타
- 0
- 포크
- 0
설치 방법
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
소스 파일 검토
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
파일 탐색기
14 개 파일SKILL.md 표시 중
SKILL.md
소스 지침 · 읽기 전용 미리보기- name
- learn-timeout
- triggers
- ["learn-timeout","timeout prediction","estimate timeout","predict timeout","timeout model"]
- description
- General-purpose timeout estimation skill. Trains dual models (duration regression + risk classification) from corpus data and observation feedback. Returns calibrated timeout predictions with confidence intervals for any task type.
- provides
- ["learn-timeout"]
- composes
- ["task-monitor","agentic-evals"]
- disciplines
- ["ml-training","observability-operations"]
# Learn Timeout
General-purpose timeout estimation that replaces fragmented Ridge/Logistic models with a unified GradientBoosting-based predictor.
## Commands
```bash
./run.sh collect # Gather training data from all sources
./run.sh train # Train both models
./run.sh predict '{"task_type":"pdf_extraction","page_count":400}'
./run.sh observe --task-id X --actual-seconds Y
./run.sh status # Model health dashboard
./run.sh benchmark # Classifier-lab backbone comparison
```
## Prediction Output
```json
{
"estimated_seconds": 4200,
"confidence_interval": [2800, 6300],
"risk_probability": 0.35,
"risk_label": "medium",
"recommended_timeout_seconds": 6300,
"features_used": ["page_count", "table_pages", "domain"],
"model_version": "2026-02-13_v1",
"duration_model_available": true,
"risk_model_available": true
}
```
## Task Types
| task_type | Key Features |
|-----------|-------------|
| `pdf_extraction` | page_count, tables, figures, file_size, domain |
| `llm_api_call` | prompt_tokens, model, provider, image_count |
| `subprocess` | command_type, input_size, complexity_hints |
| `remediation` | issue_count, issue_severity, skill_name |
## Training Data Sources
- Corpus `profile.json` + `timings.jsonl` (S00 features + actual durations)
- Supervisor run logs (`extract_timeout` events)
- Aggregate reports (extraction timing events)
- Observation feedback loop (`data/observations.jsonl`)
GitHub에서 보기