- Home
- Skills
- Agents & Orchestration
- Routing Harness
Works with the AI Agents you already use
Routing Harness
Adaptive AI workers, reasoning levels, advisor checks, and honest usage reporting.
$5
Routing Harness
Example session with this skill installed
Use Routing Harness to plan three jobs: find an exact symbol, trace a bounded execution path, and implement a cross-module feature with unclear requirements. Explain worker and reasoning choices and acceptance checks. Planning only: do not spawn agents or edit files.
- Read your context and instructions
- Compiled the routing harness
Routing plan
Planning-only example: no agents were dispatched and no files were changed. Model choices below are role requirements, not claims of observed execution.
| Job | Proposed worker | Reasoning policy | Acceptance evidence |
|---|---|---|---|
| Find an exact symbol | Fast available worker | Low, if supported | Verify definitions and call sites in source and return file references. |
| Trace a bounded execution path | Fast available worker | Medium, if supported | Verify the entry point, calls, relevant branches, and state transitions. |
| Implement an unclear cross-module feature | Capable available worker | Medium or higher, if supported | Resolve unclear requirements before editing; define file ownership and relevant behavior checks. |
Host adaptation
Map the proposed roles to the user's available models and supported settings before dispatch. If only fixed-model subagents are available, record the inherited configuration. If there are no subagents, the lead executes sequentially and reports that independent worker/advisor execution is unavailable.
Acceptance and escalation
Failed checks block acceptance. Missing context may justify one corrected fast-worker attempt. Implementation mistakes escalate to a capable worker where available. The same failure twice triggers advisor review, or a clearly labeled lead reassessment if independent advice is unavailable.
Activity and usage
Router consultations
0. Advisor consultations:
0. Worker dispatches:
0. Tokens and cache hits: unavailable for this planning-only example. No savings or cache reuse is assumed.
Next step
Clarify the feature requirements and confirm available host capabilities before implementation.
Connects securely to your tools. The creator never sees your data.
About this skill
Give each task the right worker
Routing Harness is an opt-in SKILL.md workflow for adaptive routing, worker contracts, advisor review, and a local usage dashboard. It uses your available models and supported reasoning controls.
Route and verify
Assign fast workers to bounded tasks and capable workers to ambiguous implementation. Give each worker clear boundaries and checks. The lead verifies evidence before acceptance; repeated failures escalate to advisor review where available. Hosts without subagents can follow the workflow sequentially.
Local dashboard included
Generate an offline HTML dashboard with filterable, sortable task and agent tables. Export tasks, agents, and request-level measurements as CSV and JSON. View advisor calls and triggers, retries, overrides, repairs, requested versus observed settings, cache reads/writes, and token coverage.
Read-only adapters support selected Codex and Claude Code JSONL formats plus supported OpenCode exports and SQLite schemas. Adapters have automated fixture tests and read-only checks against locally installed hosts. Formats can change, so unsupported records produce warnings.
Honest accounting
Task and advisor attribution requires a recorded ledger or explicit session bindings. Missing telemetry is unavailable, partial totals show coverage, and cache/reasoning tokens are not added twice. There are no promised savings, shared-cache guarantees, or API-price-to-subscription conversions.
Requirements
Core workflow: a compatible SKILL.md host. Optional dashboard: Python 3.10+ and local command execution, with access to explicitly selected telemetry files and an output folder. No third-party Python packages, external analytics, hosted server, or background service is required. A manually started refresh loop is optional. The HTML report works offline.
Privacy and activation
The reader exports usage metadata rather than transcript prompts or answers. Reports may contain private task names and should be reviewed before sharing. No report is uploaded automatically. Invoke Routing Harness by name using your agent's supported skill mechanism.
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.
Requires host support for subagents or multi-model dispatch. Local dashboard requires Python 3.10+. Metrics availability depends on host telemetry exposure.
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
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 28 days ago
Nothing to install. It runs in your AI agent.
- Optional
Python 3.10+Terminal access
Both are only for the optional dashboard.
Permissions
Core routing requires no terminal or file permissions. The optional dashboard uses local execution to read explicitly selected host logs/exports or a read-only OpenCode database, and writes HTML/CSV/JSON to your chosen folder. It does not crawl all sessions, export prompts/answers, transmit telemetry, or install a service. Other shell/file use depends on the coding task.