Hyphenomenon

A new Hyphenomenon project can start with a simple operator prompt: run the Hyphenomenon project-intake skill, give it the local repo path, the project title, and any live or canonical routes, then let the workflow assemble the first record. The point is speed with traceability. The operator does not manually create every project page, note, workflow, screenshot, and graph relationship from scratch; the intake loop creates a reviewable package and shows what it plans to write before anything public is trusted.

The intake skill scaffolds a dossier, screenshot folder, hydration candidates, a project-page packet, and a workflow packet when the project depends on process evidence. For this site, those artifacts matter as much as the final page: they show the repo paths that were read, the screenshots captured, the source documents hydrated into notes, the workflow steps proposed, and the dry-run reports used to decide whether the local database should change.

Automation collects and structures evidence; it does not declare the record finished. Dry-runs expose the project sections, notes, data models, workflow steps, media, edges, and overwrite risks before apply. Human review decides whether to accept the local write, revise weak copy, append evidence only, or stop. That boundary is the operating model: fast project creation, but with an explicit editorial gate.

The finished output is not a polished portfolio entry. It is a public logbook: the project page gives the narrative spine, notes preserve source-backed proof, workflows show how the work moved, screenshots show surfaces, and the Explore map shows relationships. When the process works, the archive can keep up with new work because the repeatable intake loop carries the heavy setup while the operator keeps judgment in the loop.

The September 1 source-durability work and September 8 report review exposed separate obligations: retain and restore exact sources, establish which machines and intervals a report covers, and verify the publication target. Project 1157 could read accessible source roots; that did not retrieve another machine’s data or implement cross-host aggregation. Local implementation checks therefore did not establish operational cutover or combined-machine coverage.

The September 8 review records the local daily Skysight publisher paused while collection continued, with the weekly scheduler missing on the inspected Mac and the other machine’s scheduler state unverified. It identifies publication through September 4 and a prepared September 5 report held at authentication. These are dated operational findings; no publisher or source archive was changed by this intake. A later archive-integrity finding remains an unresolved caveat rather than a repair claim.

The durable lesson is to retain explicit coverage and missing-source labels, deduplicate any attributed host imports, and reconcile the exact report before resuming publication. Multiple local roots, an empty remote completion or a prepared artifact are insufficient proof of complete, recoverable, published reporting.