Agent OS · Component · governed-agent-execution
Durable governed agent execution
What it does
Resolve skills/context/tools into execution manifests, claim work with leases and coordinate retries, resumes, human input and output fulfillment.
Core pattern
Authorized task-chain execution + retries and pauses
- Input
- Task graph and execution context
- Output
- Execution steps, outputs, traces and resumable status
Key parameters
| Name | Type | Default | Values |
|---|---|---|---|
| policy | — | — | — |
| allowed_tools | array | [] | — |
| max_steps | integer | 50 | 1–10000 |
| max_retries | integer | 3 | 0–20 |
| resume_input | — | — | — |
Maturity and limits
Execution/lease/reaper/resume tests and explicit manifest/worker implementation. Source paths were inspected statically; tests and live integrations were not executed. This rating describes implemented code, not production validation.
No root reuse license found; third-party component notices apply. Workspace data, provider credentials and local-runtime access require tenant isolation; live services were not exercised.
Used in pipelines
Industry tags
- Manufacturing & industryTransferable use
For maintenance work-order assignment and handoff, use authorized task-chain execution followed by retries and pauses.
Skills pack
Reference pack with offline tests
Download · 18KBSHA-256 cae043621a16e44827662a909b4250e011abb734b7aad8a0a15b5cae29e25c97
Related components
-
Recoverable agent runtime and context management
Durable conversation events + budgeted checkpoint resumption
Working 1 pipeline -
Goal planning and versioned task graphs
Goal decomposition and dependencies + versioned revisions
Working 1 pipeline -
Persistent goals, bounded subagents and scheduled tasks
Capability-scoped delegation + concurrency and schedule limits
Working 1 pipeline -
Reusable skill extraction, evaluation and solution packs
Completed artifacts to task packs + version evaluation
Working 1 pipeline