Three evidence records
Can each workflow be reproduced and checked?
Each record names the job, selected context, runtime owner, input, expected review, screenshot, and limitation.
Turn launch notes into reviewable next actions
- Job
- Convert a local collection of launch observations into a short next-action list.
- Selected context
- Select the relevant BenAgent Note and, when useful, a Prompt Inventory instruction that defines the required output format.
- Runtime
- Hermes or OpenClaw endpoint configured by the user
- Reproducible input
- “Organize these launch notes into the next three actions. For each action, state the owner information that is still missing. Do not invent dates or owners.”
- BenAgent role
- Prepare the selected context, submit the bounded request, show progress and retain the returned result in the local conversation.
- Human review
- Check every action against the original note, fill missing owners and reject invented commitments.
- Limitation
- This workflow does not prove measured time savings and does not grant the runtime access to unselected project systems.
Turn selected clipboard material into a bounded draft
- Job
- Reuse a copied excerpt or image without silently uploading the entire Clipboard History.
- Selected context
- Enable opt-in Clipboard History, choose one relevant entry, use Send to Agent and review the prepared composer content before submission.
- Runtime
- Hermes or OpenClaw endpoint configured by the user
- Reproducible input
- “Summarize the selected material in five bullets. Separate statements supported by the material from open questions.”
- BenAgent role
- Keep the history local until selection, prepare the selected content, submit only after review and retain the returned draft locally.
- Human review
- Compare each bullet with the selected source and remove unsupported conclusions.
- Limitation
- A file reference copied into Clipboard History is sent as text; BenAgent does not automatically read and upload the referenced file bytes.
Receive a supported background result on the Mac
- Job
- Surface a useful runtime result without keeping the runtime terminal in the foreground.
- Selected context
- Configure the repo-local Hermes or OpenClaw delivery plugin and target BenAgent's loopback-only event listener.
- Runtime
- The configured runtime owns and executes the cron, hook or plugin work
- Reproducible input
- “Create a runtime-owned test job that returns a descriptive final result and uses the supported BenAgent delivery path.”
- BenAgent role
- Accept a supported loopback event, deduplicate it, queue it when needed and present it through native result surfaces.
- Human review
- Confirm the event source, job identity and returned content before acting on the result.
- Limitation
- BenAgent does not create a native durable schedule object or guarantee delivery for every runtime job. The runtime and plugin own execution and retries.
Reproduction protocol
How should a reviewer repeat these workflows?
- 01
Install the current App Store build
Record the BenAgent version, macOS version, runtime, endpoint location, and model provider before the test.
- 02
Use non-sensitive test context
Create or select only the record named by the workflow. Do not grant unrelated file or account access.
- 03
Capture the full ownership path
Record the prepared handoff, endpoint readiness, visible progress, final result, and whether a plugin path was required.
- 04
Publish failures and limits
Report empty replies, authentication failures, unsupported progress, provider-side retention, and manual corrections alongside the successful path.
- No unpublished capability should be inferred from a screenshot.
- No productivity gain should be quantified without a study.
- No affiliation or provided access should be hidden in an independent review.
Frequently asked questions
Questions about these evidence records
Do these workflows prove that BenAgent saves time?
No. They are reproducible product-evidence records, not a productivity study. They show the current handoff, runtime, result, review, and limitation path without claiming a measured time-saving percentage.
Does BenAgent choose and upload all available context automatically?
No. The user chooses relevant Notes, Prompt Inventory entries, Clipboard History items, images, or conversation context and reviews the prepared request before submission.
Who is responsible for the result?
The configured runtime and model provider generate the result under their own capabilities and terms. The user remains responsible for checking the result before using it, publishing it, or taking an external action.
Can every runtime job send a background result to BenAgent?
No. Background delivery requires a supported Hermes or OpenClaw plugin path and a reachable BenAgent loopback listener. The runtime owns execution and plugin retries; BenAgent owns only accepted local delivery and presentation.