agentsclimarketplace

Omega3 distill

Skill mohamed-essam/omega3-template/skills/omega3-distill

Nightly "sleep on it" pass for the omega3 knowledge hub. Drains inbox into systems/knowledge (dedup + link), optionally refreshes a knowledge graph, updates per-initiative journal rollups, and does a review-gated generative pass proposing new cross-system insights. Meant to run autonomously on a nightly schedule (e.g. cron) as `claude -p "/omega3-distill"`; can be run manually.From its SKILL.md

Install
npx -y skills add mohamed-essam/omega3-template --skill omega3-distill

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 29 days oldThe repository was created 29 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

4.7 KB, ~1.1k tokens by cl100k_base, as published. Nobody here has run it

omega3-distill

Meant to run autonomously (nightly, headless — see the repo README for cron/scheduler setup). Be conservative and auditable: never delete curated notes, never auto-approve insights, honor redaction, and log everything you do. OMEGA3=__OMEGA3_HOME__.

0. Setup

cd __OMEGA3_HOME__
git pull --rebase 2>/dev/null || true

Read .omega/last-distill — the ISO timestamp bounding "new since last run". Keep running counters for the digest (§8).

1. Drain inbox/ → systems/ + knowledge/ (dedup + merge)

For each inbox/*.md (skip README.md and the proposals/ subdir):

  1. Read it; determine target folder from type: (system → systems/, knowledge → knowledge/).
  2. Look for an existing note on the same subject (match by slug, title, or systems: tag; also try graphify query if you use the graph).
    • Exists → merge new info in, refresh updated:, add any missing [[links]]. Don't duplicate.
    • New → create from the matching templates/ file, filled out, with [[links]] to related notes.
  3. Update the destination folder's INDEX.md (one dense line under its marker comment).
  4. git rm/delete the drained inbox file. Leave anything ambiguous in inbox and flag it in the digest rather than guessing.

2. Harvest per-repo session buffers (optional supplement)

If you keep per-repo session buffers (e.g. .remember/ dirs under your repos root), scan them for entries modified after last-distill and pull only durable, high-signal items the session-end skill missed; treat each like an inbox item (§1). Best-effort — skip on any error. Skip this section entirely if you don't use per-repo buffers.

3. Refresh the graph (OPTIONAL — only if you use graphify)

This hub works fully as plain markdown + [[wikilinks]]; the knowledge graph is an optional accelerator. If you use graphify, refresh it incrementally:

graphify __OMEGA3_HOME__ --update

graphify reads its LLM backend from your environment (an API key or a cloud provider profile) — configure that per graphify's own docs; nothing about it is baked into this hub. If graphify is unavailable or errors, note it in the digest and continue — the rest of distill does not depend on it. Skip this section entirely if you don't use graphify.

4. Update journal rollups

Read journal/*.md entries since last-distill plus existing journal/rollups/*.md. For each initiatives: slug seen, create/update journal/rollups/<slug>.md: a running, promotion-ready narrative — situation, my role, drivers, actions to date, and quantified impact (aggregate results across entries). Update journal/rollups/INDEX.md.

5. Generative pass — propose insights (REVIEW-GATED)

Look for cross-system patterns, relations, contradictions, and recurring themes — use graphify query over the graph (if you use it) and the notes changed this run. For each candidate insight:

  1. Skip if it's already covered by a note in insights/.
  2. Skip if it matches anything in .omega/rejected.md (by slug or fingerprint/claim).
  3. Otherwise write it to inbox/proposals/<slug>.md from templates/insight.md, filling confidence: and provenance: (which notes/edges it's built from; EXTRACTED vs INFERRED). Never write to insights/ directly — that only happens via /omega3-review. Be selective: a few high-value proposals beat many shallow ones.

6. Update indexes

Ensure every folder's INDEX.md reflects reality (added/merged/renamed notes).

7. Marker

Rewrite .omega/last-distill:

date -u +%Y-%m-%dT%H:%M:%SZ > .omega/last-distill

(keep the leading comment line).

8. Digest + commit + push

Append a dated block to .omega/distill-log.md (see its format) summarizing: inbox drained (counts + slugs), buffers harvested, graph status, rollups updated, insights proposed, and anything left unprocessed/flagged. Then:

git add -A
git commit -m "distill <YYYY-MM-DD>: <N inbox, M insights proposed, rollups K>"
git push

Guardrails

  • Idempotent-ish: re-running the same night must not create duplicates (dedup in §1/§4).
  • Never delete or overwrite curated systems//knowledge//insights/ content wholesale — merge additively.
  • Honor the redaction rule on everything you write.
  • If the working tree has unexpected uncommitted user edits, commit them separately or leave them — never discard.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Keep looking

Skills are one crate of 326,645. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.