Agent Operating Guide
/coach-workflow-scan
A read-only triage of the steps assigned to you, sorting each into ready, needs-input, blocked, or already-handled with a one-line reason.
Triage the workflow steps assigned to you. The skill reads your assigned steps and sorts each into one of four states, ready-to-execute, needs-input, waiting-on-upstream, or already-handled, plus an awaiting-your-approval line for steps you approve, each with a one-line reason. It’s read-only: a what’s-on-my-plate brief that never changes anything.
When to use
Section titled “When to use”When you want to know what to work on next, triage your queue, or get a morning brief of your plate. It also runs as the scan phase of an autopilot pass. To act on the ready steps, follow it with /coach-workflow-execute.
Inputs
Section titled “Inputs”| Flag | Required | Notes |
|---|---|---|
--limit <n> |
No | Max assigned steps to pull. Default 50. |
--include-approver |
No | Include the awaiting-your-approval and approver-waiting rows. Default on. |
--headless |
No | Suppress the interactive handoff and return the brief only. |
Example
Section titled “Example”/coach-workflow-scan pulls your assigned steps and classifies each: checking whether
it’s already handled, whether its upstream is clean, and whether it still needs an input
only you can supply. It returns a brief grouped by bucket: “Queue scan: 3 ready, 2 need
you, 1 blocked, 4 already handled.” Each row names its reason, e.g. “waiting on Draft
findings (owner: Alice).”
Good to know
Section titled “Good to know”- Four doer-states, most-specific-wins. Already-handled (has a result or a prior automated suggestion) is checked first, then waiting-on-upstream (a direct upstream step isn’t COMPLETED or APPROVED), then needs-input (a decision or document only you can supply), then ready-to-execute (upstream clean, full context in hand).
INPUT NEEDEDplaceholders don’t count as handled. A pending autopilot request for a missing input keeps its step live: the scan re-checks it every run, marks it ready the moment the input arrives, and keeps it in needs-input (without re-requesting) while the input is still missing.- Rubber-stamped placeholders are flagged. An upstream step approved while it still
contained the
INPUT NEEDEDplaceholder blocks its downstream steps and is listed under “Approved placeholders, reopen these”, placeholder text is never treated as usable upstream content. - Judgment steps land in needs-input. If a step needs your decision, that decision is the missing input, so it’s never mislabeled ready-to-execute.
- Read-only. Scan never drafts or submits; the acting is /coach-workflow-execute’s job, and the scheduling loop is autopilot’s.
- Standalone morning brief. Run it by hand any time to see your plate; the same scan is the first phase of an autopilot run.
Related
Section titled “Related”- /coach-workflow-execute: execute the steps this marks ready.
- /coach-workflow-monitor: the supervisory sweep of everyone’s steps.
- Builds on get_current_context, get_step_context, and query_data: all read-only.