edge-verifier
Verifies relationships between a source text (e.g. Episode, Task) and existing Knowledge Graph lessons. Runs KNN/Hybrid search to find candidates, then uses LLM (scillm) to verify "verifies", "contradicts", or "related" stances with strict rationales.
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
8 개 파일SKILL.md 표시 중
SKILL.md
소스 지침 · 읽기 전용 미리보기- name
- edge-verifier
- description
- Verifies relationships between a source text (e.g. Episode, Task) and existing Knowledge Graph lessons. Runs KNN/Hybrid search to find candidates, then uses LLM (scillm) to verify "verifies", "contradicts", or "related" stances with strict rationales.
- internal
- true
- allowed-tools
- Bash
- triggers
- ["verify edges","link content","generate relationships","schedule verification"]
- metadata
- {"short-description":"KNN + LLM verification for KG edges"}
- provides
- ["edge-verification"]
- composes
- ["scillm","task-monitor","agentic-evals"]
- taxonomy
- ["validation","knowledge","precision"]
- disciplines
- ["memory-knowledge","evaluation-quality"]
# Edge Verifier
This skill links new content to the existing Knowledge Graph by:
1. **Recall**: Running a KNN/Hybrid search (`graph_memory.search`) to find potential related lessons.
2. **Verify**: Using `scillm` (LLM) to audit the relationship, classifying it as `verifies`, `contradicts`, or `neutral`.
3. **Link**: Creating verified `lesson_edges` in ArangoDB.
## Usage
### Manual Execution
```bash
# Verify edges for a specific source text
.pi/skills/edge-verifier/run.sh --source_id <ID> --text "Content to verify..."
# With explicit options
.pi/skills/edge-verifier/run.sh --source_id <ID> --text "..." --type "episode_turn"
```
## Scheduling (Scheduler Integration)
This skill is designed to be scheduled via the `/scheduler` skill for continuous verification of new memory artifacts.
### Register with Scheduler
```bash
.pi/skills/scheduler/run.sh register \
--name "edge-verify-nightly" \
--cron "0 2 * * *" \
--command ".pi/skills/edge-verifier/run.sh --batch" \
--description "Nightly edge verification"
```
Jobs and logs are stored under `~/.pi/scheduler/`. See `.pi/skills/scheduler/SKILL.md` for full options.
## Common Mistakes
### WRONG: Running verification without ArangoDB credentials
```bash
./run.sh --source_id <ID> --text "content" # fails silently without .env
```
### RIGHT: Ensure .env has ArangoDB and Chutes credentials
```bash
# Verify .env exists with ARANGO_URL, ARANGO_DB, CHUTES_API_KEY
cat .env | grep -E "ARANGO|CHUTES"
./run.sh --source_id <ID> --text "content"
```
### WRONG: Verifying edges without specifying content type
```bash
./run.sh --source_id <ID> --text "content" # default type may not match
```
### RIGHT: Specify the source type for accurate classification
```bash
./run.sh --source_id <ID> --text "content" --type "episode_turn"
```
### WRONG: Running batch verification during peak hours
```bash
./run.sh --batch # competes with interactive queries for LLM slots
```
### RIGHT: Schedule batch verification for off-peak (nightly)
```bash
.pi/skills/scheduler/run.sh register --name "edge-verify-nightly" \
--cron "0 2 * * *" --command ".pi/skills/edge-verifier/run.sh --batch"
```
## Prerequisites
- `.env` must expose ArangoDB credentials.
- `CHUTES_API_KEY` for LLM calls.
GitHub에서 보기