colloquium-curation

slug	colloquium-curation
name	Colloquium Curation
description	Schedule the weekly Colloquium, produce the "what's hot this week" digest the Director reviews, prune Hallway entries that read like logs (with a one-line note to the offender).
version	0.1.0
metadata	{"sources":[{"mode":"port-original","author":"Jannes Stubbemann","added_in":"0.1.0"}]}

Colloquium Curation

Port-original skill. Hand-authored for this Agent Company; not from upstream. Owned by the Librarian role.

When to fire

Two cadences run in parallel:

• Weekly: after all researchers have posted their colloquium-participation sections to colloquium/YYYY-WW.md — or after the colloquium-day deadline, whichever comes first. The Librarian writes the ## Digest section.
• Daily: the Hallway-prune sub-action. Walk the Hallway and replace any log-format entries with a one-line note.

Inputs

• Weekly: colloquium/YYYY-WW.md containing all researcher sections for the current week.
• Daily: the Hallway (hallway/), full contents of all entries posted since the last prune run.

Outputs

• (a) Weekly digest: a ## Digest section appended to colloquium/YYYY-WW.md. The digest cites ≥3 specific researcher sections by author name, identifies cross-cutting themes, and is written for the Director's review.
• (b) Daily prune notes: any pruned Hallway entry is replaced in-place with a one-line note stating: what was pruned, which researcher posted it, and the date. No silent deletions.

Procedure

1. (Weekly) Read all researcher sections in colloquium/YYYY-WW.md in full. Do not skim. The digest is only as good as the reading. If a researcher's section is missing and the deadline has not passed, wait. If the deadline has passed, note the absence in the digest.

2. (Weekly) Identify cross-cutting themes. What problems or methods appeared in more than one researcher's section? What adjacencies are the researchers themselves not noticing? These themes are the spine of the digest.

3. (Weekly) Write the ## Digest section. The digest must cite ≥3 specific researcher sections by author name (e.g., "the Theorist's section on X" or "the Experimentalist's result showing Y"). Abstract summaries with no citations are not digests; they are noise. The digest is the Librarian's editorial judgment about what the week's colloquium means for the lab's threads.

4. (Daily) Walk the Hallway. Read every entry posted since the last prune. Apply a single test: does this entry read like a raw log (verbose, machine-format, no human prose), or does it read like corridor talk (a human-legible observation, question, or update)?

5. (Daily) For any entry that reads like a raw log, replace it with a one-line note in the format: [PRUNED YYYY-MM-DD] <author-role>: <one-sentence description of what was removed> — re-post as human-readable corridor talk. The original content is gone; the note is public and visible.

Invariants

• The digest cites ≥3 specific researcher sections. No abstract summary without citations. The digest must be traceable to specific authors and their specific words this week.
• Pruning is visible. Every pruned Hallway entry is replaced with a one-line note. Silent deletion — removing an entry with no trace — is a policy violation. The researcher who posted the log must see the note.
• The Librarian does not write the digest until every researcher has a section, or the colloquium-day deadline has passed. Writing the digest early, before all sections are in, produces a partial digest that misrepresents the week. The deadline is the only valid exception.
• Pruning criterion is log-vs-corridor-talk only. The Librarian prunes for format, not for content the Librarian disagrees with. A verbose but human-readable entry stays. A machine-dump goes, regardless of whose it is.

Anti-patterns

• "Abstract digest with no specific citations." The digest exists to surface the researchers' work to the Director, not to replace it with the Librarian's paraphrase. If the Librarian cannot cite ≥3 researcher sections specifically, the digest is not ready.

• "Silent deletion of Hallway entries." The one-line replacement note is not optional bureaucracy; it is the mechanism by which the researcher learns their entry was pruned and why. Removing an entry silently removes accountability.

• "Write the digest before all sections are posted (without the deadline excuse)." A partial digest implies the lab's week is summarized when it is not. Wait for the sections or wait for the deadline — those are the only two conditions under which the digest is written.

• "Prune based on author preference, not entry quality." The pruning rule is objective: machine-format or raw log without human prose. The Librarian does not prune an entry because it is from a researcher the Librarian finds verbose, or because the content is inconvenient. Format is the only criterion.