deploy-agentic-rag — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited deploy-agentic-rag (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
| Aspect | Standard RAG (Play 01) | Agentic RAG (Play 21) |
|---|---|---|
| Search decision | Always search | Agent decides IF needed |
| Source selection | Fixed single source | Agent picks from multiple |
| Iteration | One-shot | Agent iterates if insufficient |
| Self-evaluation | None | Agent checks groundedness |
| Multi-source | No | AI Search + Bing + SQL + APIs |
az account showaz bicep build -f infra/main.bicep
az deployment group create -g $RG -f infra/main.bicep -p infra/parameters.json{
"sources": [
{ "name": "knowledge-base", "type": "ai-search", "priority": 1 },
{ "name": "web-search", "type": "bing", "priority": 2 },
{ "name": "product-db", "type": "sql", "priority": 3 }
],
"routing": { "default_sources": ["knowledge-base"], "max_sources_per_query": 3 }
}The agent has a tool per data source. It autonomously decides which to call:
search_knowledge_base — internal KB docssearch_web — web for recent infoquery_database — product/customer dataself_evaluate — check if context sufficient| Parameter | Default | Range | Purpose |
|---|---|---|---|
max_retrieval_hops | 3 | 1-5 | Max search iterations |
min_groundedness | 0.85 | 0.7-0.95 | Self-eval stop threshold |
max_sources | 3 | 1-5 | Max sources per query |
iteration_timeout | 30s | 10-60s | Total retrieval budget |
| Issue | Cause | Fix |
|---|---|---|
| Agent searches all sources | No routing logic | Improve source descriptions in tools |
| Infinite loop | No iteration cap | Set max_retrieval_hops=3 |
| Cache never hits | Threshold 0.99 | Lower to 0.92 |
| Slow (>30s) | Too many iterations | Reduce max_hops, raise groundedness |
| Citations missing | Source not in tool output | Return source metadata from tools |
| Wrong source selected | Ambiguous tool descriptions | Make tool descriptions more specific |
- name: Validate Infrastructure
run: az bicep lint -f infra/main.bicep
- name: Deploy to Staging
run: az deployment group create -g $STAGING_RG -f infra/main.bicep -p infra/parameters.json
- name: Run Retrieval Quality Gate
run: python evaluation/eval.py --metrics source_selection,groundedness --ci-gate
- name: Verify Source Connectivity
run: python scripts/test_sources.py --sources knowledge-base,web-search,product-dbUser Query → Agent (gpt-4o with tool-calling)
│
├─ Tool: search_knowledge_base → AI Search
├─ Tool: search_web → Bing API
├─ Tool: query_database → SQL
├─ Tool: call_api → Custom REST
└─ Tool: self_evaluate → Check groundedness
│
├─ Sufficient? → Respond with citations
└─ Insufficient? → Iterate (next source or refine query)~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.