- Home
- Skills
- Agents & Orchestration
- DAG orchestrator skill
More screenshots
Works with the AI Agents you already use
DAG orchestrator skill
Enterprise multi-agent DAG orchestrator with 4-tier epistemic conflict resolution and cryptographic provenance.
Free
DAG orchestrator skill
Example session with this skill installed
Does ACME exposure breach the risk threshold, and does the pricing guardrail allow the re-price? Cite the policies.
- Read your context and instructions
- Compiled the dag orchestrator
I have processed the compound analytical query by compiling a 3-wave execution DAG and reconciling data across the Tier 1 compute and Tier 2 policy engines.
Orchestrator Synthesis
Request ID: req-8829-xf
State Fingerprint: 8f2d93a1...
| Component | Status | Details |
|---|---|---|
| Tier 1 Compute | OK | Current ACME Exposure: $4.2M. |
| Tier 2 Policy | OK | Rule 57 (Risk Threshold): $5.0M. Rule 112 (Pricing): Allowed. |
| Conflict Resolution | Resolved | Tier 1 compute metrics prioritized over Tier 3 retrieval variance. |
| Provenance | Verified | All citations matched vector_chunk_sha256 in the Protected 50 corpus. |
Triggered Policies
- Policy 1.2.4 (Quantitative Risk): Exposure of $4.2M is within the $5.0M threshold. Verdict: PASS.
Policy 5.1.0 (Pricing Guardrail): Entity ACME is eligible for re-pricing under active identity ACTIVE_200. Verdict:
ALLOW.
Telemetry Summary
- Total Latency: 1240ms (p50 target met).
- Execution Waves: 3.
- Budget Slack: 1360ms.
Next steps
- Review the
provenance_ledgerfor specific document IDs associated with Rule 57. - Check the
slm_shadow_telemetrylogs for intent agreement rates. - Proceed with the re-pricing workflow as authorized by the symbolic AST engine.
Connects securely to your tools. The creator never sees your data.
About this skill
Complex analytical queries often fail when they require data from multiple conflicting sources or exceed latency budgets. This skill provides a high-assurance orchestrator that compiles multi-hop queries into validated Directed Acyclic Graphs (DAGs) while enforcing strict enterprise guardrails.
What it does
- DAG Compilation transforms complex tasks into parallel execution waves using Kahn's algorithm to prevent cycles.
- Epistemic Reconciliation resolves conflicts between different data sources using a 4-tier hierarchy where deterministic compute always overrides unstructured retrieval.
- Provenance Verification performs cryptographic SHA-256 checks on every retrieved data chunk to prevent spoofing or unauthorized data access.
- SLA Enforcement prunes optional low-priority tasks to keep end-to-end latency within defined p95 and p99 millisecond targets.
- Context Compression uses a knapsack algorithm to fit the most relevant facts into a 4096-token window while pinning high-priority data.
How it works
- Intent Routing detects compound multi-hop analysis requests and triggers the DAG planner.
- Task Decomposition partitions the query into discrete nodes with assigned priorities, timeouts, and dependencies.
- Parallel Execution dispatches subtasks across MCP tools with circuit breakers and adaptive concurrency limits.
- Conflict Synthesis merges results, applying the epistemic hierarchy to discard low-confidence or contradictory claims.
Frameworks & tools
Built with Python 3.11 using asyncio.TaskGroup and Pydantic v2 for frozen data contracts. It integrates with external systems via MCP, targeting Qdrant, PostgreSQL, and vLLM backends.
Why this beats prompting it yourself
Standard prompts cannot guarantee cycle-free task execution or enforce millisecond-level latency budgets across multiple tool calls. This skill replaces unreliable "chain of thought" with a structural execution engine that handles circuit breaking, hedging, and cryptographic provenance checks that prompts alone cannot perform.
Use cases
- Risk Exposure Analysis calculating breach thresholds by synthesizing real-time financial metrics with static regulatory policies.
- Conflict Resolution reconciling disagreeing facts between a vector database and a symbolic policy engine.
- High-Throughput Analytics managing 60+ subtasks simultaneously without exceeding API rate limits or context windows.
Known limitations
Currently prohibited from production connections pending Architecture Review Board audit. Uses a synthetic 550-identity corpus and in-process reference transports instead of live database drivers.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
The vector provenance digest strictly covers chunk text rather than metadata, and natural language prompt-injection screening must be handled at your API gateway.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 7 days ago
- Free with an Agensi account
- Software
Python
Permissions
File Scopes
This skill enforces a hard 3000ms execution SLA and requires an environment capable of running asynchronous Python 3.11+ tasks.