Generated from
folio-assistant-core/processes/content/review-log-watch.bpmnbygen-processes-viz.ts— do not edit here. All processes · folio-assistant-core processes
Review-log watch
Process_ReviewLogWatch · strict (defaulted) · 7 step(s)
One poll of one review log (a drafting Contributor Feedback sheet, or a
public-comment sheet’s Review Change Sets) and its folio repository. A tool decides, with no agent judgement, whether
anything a person set in the sheet or any open issue changed since the
last poll. Only then does the agent append rows for new issues and do
its pass. Skill: review-log-watch.
How it connects
- Called by: no call activity names this process
- Calls: Contributor feedback on a working draft, Record agent step
- Names the
review-log-watchskill without calling this process: Record agent step —activity-calls-skill-processasks whether each should be a call activity. - Presented on: no docs page section shows this diagram
- Skill:
review-log-watch - Running instances: none recorded
Lanes — who acts
| lane | role | what it does here |
|---|---|---|
| Log watcher (authoring agent) | authoring-agent | The agent thread that watches the log. Until the check says changed it only reads and runs the tool. |
Steps
Every one of the 7 step(s) is documented.
| step | lane | skill / sub-process | what it does |
|---|---|---|---|
Read the Contributor Feedback tabTask_ReadSheet | Log watcher (authoring agent) | review-log-watch | Reads Watch!F1:I1, the hash of the person-set cells that the sheet script writes on every refresh, and Contributor Feedback’s header row and Issue column (a public-comment log: Review Change Sets’ header row and Change-set column), and saves them as JSON for the check. A sheet without the Watch hash yet is read whole. Reading is not agent work: it changes nothing. |
Run the review-log-watch checkTask_RunCheck | Log watcher (authoring agent) | review-log-watch | Runs Tool review-log-watch with the folio repository, the saved values and the log’s state file. The tool reads the repository’s open and closed issues (no pull requests, no issue labelled no-sheet) and compares a digest of them and of the cells people set (Title, Selected Sections, Contributor Feedback, Disposition, Ready for agent, Ready for merge, Dispositions confirmed; on a public-comment log also Committee Feedback, TWG (CS), TWG (human), Change Set (human)) with the last check’s. A read with no Issue column tracks only change-set issues and appends no rows, reporting them as issue-unlinked. It writes nothing. |
Hand new feedback to the row’s agent; take over stalled rowsTask_StallCheck | Log watcher (authoring agent) | review-log-watch | On every poll, changed or not, new Blue feedback goes verbatim to the agent working the row, and a stalled row is restarted or taken over. For each row: new Blue feedback goes, verbatim and in the same poll, to the thread that owns the row (named in its Agent row or Feedback (agent)), and Feedback (agent) says who has it. A row is stalled when its Blue cells changed, or Ready for agent turned Yes, more than 10 minutes ago and no Agent step was recorded since; or when its agent recorded no step for 2 hours. A stalled row’s owning thread is asked once; with no answer within 10 minutes, or a thread that is done or stopped, the watcher takes the row over, or asks the coordinator for a new thread, and writes the new owner, the step and a link into Feedback (agent). |
Append rows for new issues; hide closed onesTask_AppendRows | Log watcher (authoring agent) | review-log-watch | For each row in the tool’s append list (one per issue that no row’s Issue cell links yet), appends it below the last row of Contributor Feedback: Title is the issue’s title, the first Contributor Feedback column @login (#N): the issue’s first paragraph, Issue its URL. The script gives each new row its Change-set id. No other writer adds issue rows. Then each row linking a closed issue (the tool’s closedRows, and the rows just appended as closed) is hidden and its text struck through. |
Agent pass on the changed rowsTask_AgentPass | Log watcher (authoring agent) | calls Contributor feedback on a working draftreview-sheet-script | For each changed row, the agent pass of Contributor feedback on a working draft: Ready for agent = Yes gets its edit, Ready for merge = Yes its merge, and a Blue answer to a Decision needed question is taken up. Every new row (from an issue, a person or a TWG workbook), every row with an empty Feedback (agent) and every row whose Blue cells changed gets feedback in the same wake, in the fixed layout: the ask first (Decision needed: with lettered options, one recommended, and where to answer), then Proposed direction: and Next:, only the tags bold. The work stays inside the row’s Selected Sections; a row with none gets a question asking which, and any edit outside them is named on an Outside the selected sections: line. PRs are named as repo#n, which the sheet script links, and it adds a Where to review line linking each selected section on the preview. The agent then runs tools/review-log-watch/lint.ts and fixes what it flags: a cell that breaks the layout, or a PR state that GitHub contradicts. The agent records its step on the Agent tab at once, with the real time. Previews follow main: by default (Config PREVIEW_REFRESH automatic) the agent runs tools/review-log-watch/previews.ts on every poll for every open change-set PR, and also when people act on a change set (new Blue feedback, Ready for agent set to Yes, or feedback asking to update the preview); with PREVIEW_REFRESH on-feedback, only then. A branch behind main gets main merged in (a merge commit), pushed and restaged; a conflict a person must settle is asked in the change set’s feedback. |
Record agent step on each row it wroteTask_RecordStep | Log watcher (authoring agent) | calls Record agent stepreview-sheet-agent-columns | For every row the pass or the stall check wrote, Record agent step: the Agent tab row first (status, pull request, preview, step), then Feedback (agent) from it, then read back Agentic status and Staging review. A row whose Feedback (agent) names a pull request or preview its Agent row does not, or whose Agent step is not “<step> <time>”, is fixed here. |
Accept the new stateTask_Accept | Log watcher (authoring agent) | review-log-watch | Runs the check again with –accept, so the state file holds what was just handled and the next poll compares against it. A row the pass could not finish stays changed and is picked up on the next poll. |
Decisions
Every one of the 1 decision(s) is documented.
| decision | what decides it | branches |
|---|---|---|
Sheet or issues changed?Gateway_Changed | The tool’s status. unchanged (exit 0): the poll ends and no other agent work starts. changed (exit 10): its changes and the rows to append go on. | changed → Append rows for new issues; hide closed ones unchanged → Nothing changed |