| name | dikw-compile |
| version | 1.0.0 |
| description | KOS-flavored post-ingest compilation pass. After any ingest skill creates
a source page, walk the DIKW (Data → Information → Knowledge → Wisdom)
compilation: identify impacted concept/synthesis/decision pages, enforce
strong-link type annotations (supplements/contrasts/implements/extends),
2-5 link budget per page, and bidirectional citations.
Runs after idea-ingest / media-ingest / meeting-ingestion.
|
| triggers | ["compile this source","dikw compile","after ingest"] |
| tools | ["get_page","put_page","search","backlinks","add_link"] |
| mutating | true |
DIKW Compile Skill
Port of KOS v1 knowledge/agents/prompts/compile-agent.md. The core insight:
a meaningful source should usually update more than one page. If ingest
only produces a disconnected summary, compilation hasn't happened yet.
Contract
Guarantees on every run:
- Source page has at least one downstream link (to concept/synthesis/decision)
- Every new or updated link carries a type annotation
- Link budget per page: 2-5 strong links, not a dump of weak associations
- Back-links are symmetric (iron law, from
maintain skill)
- Compile grade is recorded in source page frontmatter (A/B/C/F)
Link types (port from KOS meta/link-rules.md)
| Type | When to use |
|---|
supplements | new evidence reinforces existing page |
contrasts | new evidence challenges or disagrees |
implements | source is concrete realization of abstract concept |
extends | source adds scope/depth previously missing |
Do NOT create links of type "see also" or "related" — that's the
weak-link anti-pattern KOS v1 explicitly banned.
Protocol (6 phases)
Phase 1 — Impact analysis
Given new source page <source-slug>:
gbrain query "<source-title-keywords>" --no-expand | head -20
Collect candidate impact pages (>= 0.5 similarity).
Phase 2 — Classify candidates
For each candidate, read its compiled truth. Decide:
- Is the source new evidence for an existing claim? →
supplements
- Does it challenge an existing claim? →
contrasts
- Is it a concrete instance of an abstract framework? →
implements
- Does it open a new facet not covered before? →
extends
- If none of the above, skip. Don't force a weak link.
Phase 3 — Enforce budget
Sort candidates by strength. Keep top 5. If you can't find 2 strong links,
the source may not be compilation-worthy yet; flag for later re-try rather
than manufacture weak links.
Phase 4 — Write links
gbrain link <source-slug> <target-slug> --type <supplements|contrasts|implements|extends>
Update target page's See Also or equivalent section with inverse reference,
preserving citation format from skills/conventions/quality.md.
Phase 5 — Grade
Grade this compilation event and write compile_grade to source frontmatter:
| Grade | Condition |
|---|
| A | 3+ targets updated AND at least 1 new synthesis/decision created |
| B | 1-2 targets updated with precise strong links |
| C | Only source page written, no downstream updates |
| F | Source page itself fails evidence-gate check |
Phase 6 — Propose writebacks
If the source contains a reusable insight spanning multiple existing pages,
propose a new synthesis or comparison page. Do NOT auto-create; emit
the proposal to the user (or to the briefing cron for daily review).
Output
Append to ~/brain/log.md:
<YYYY-MM-DD> | dikw-compile | <source-slug> | grade=<A|B|C|F> | links=[<type>:<target>, ...]
Delegation
- Citation format:
skills/conventions/quality.md
- Back-link iron law:
skills/maintain/SKILL.md
- Evidence level prerequisite:
skills/kos-jarvis/evidence-gate/SKILL.md
- Type mapping (where new synthesis/decision lives in GBrain dirs):
skills/kos-jarvis/type-mapping.md