| name | applied-research-craft |
| description | Use when authoring or revising any canonical, cursor rule, skill, or decision-register row in this AKOS workspace and the framing introduces a novel position. Codifies the internal-evidence-sweep + external-research-sweep + citation-integration + wave-closure-research-enrichment rhythm per akos-applied-research-discipline.mdc RULES 1-3. Triggers on research grounding, applied research, external citation, novel framing, internal precedent, evidence sweep + research sweep, RESEARCH_HEAD_DISCIPLINE, wave-close research enrichment, citation integration. Pairs with .cursor/rules/akos-applied-research-discipline.mdc (the WHEN); this skill is the HOW. |
| metadata | {"author":"FraysaXII"} |
Applied-Research Craft
Codified at I86 Wave R close (drain7) from the operator's framing at the I86 Wave H Combo C+D Lane D ratify gate (2026-05-19): "applied research is a hot topic today and a competitive advantage if properly industrialized". The craft turns research grounding from afterthought ornament into compounding doctrine. Ratified by D-IH-86-CT alongside the parent rule.
When to use this skill
Read this skill before:
- Authoring a new canonical CSV (RULE 1 — internal + external research backing mandatory).
- Authoring a doctrine canonical, cursor rule, skill, or decision-register row that introduces a novel framing (RULE 2 — external citation mandatory).
- Closing a wave inside a cluster-coordinator initiative (RULE 3 — "Research enrichment" subsection in wave-closure report).
- Triaging a research need surfaced by an inline-ratify gate Principle 1.5 trigger.
This skill assumes you have already read akos-applied-research-discipline.mdc (the when) + RESEARCH_HEAD_DISCIPLINE.md (the parent doctrine).
Core principles
Principle 1 — Distinguish evidence-sweep from research-sweep
These are two different sweeps with different tools:
| Sweep | Tools | What it searches | When it fires |
|---|
| Evidence-sweep | Grep, Read, Glob, SemanticSearch | Repo state (canonicals, decision rows, code, prior planning) | EVERY canonical/rule/skill mint (per akos-inline-ratification.mdc Principle 1) |
| Research-sweep | WebSearch, WebFetch, optional reads of cited corpus | Industry precedent (academic papers, vendor docs, industry analysts, named books) | Conditional on novelty (per RULE 2 of parent rule) |
They are NOT substitutes. Novel framings require BOTH. Evidence-sweep alone surfaces "the repo is silent on this" — that's the signal to invoke research-sweep, not the conclusion that grounding isn't needed.
Principle 2 — Cite inline, not in a separate dossier
External citations belong in the canonical body or its Evidence-base section, NOT in a separate research-dossier file. Inline citation is half of the load-bearing claim — the doctrine says "X because Y (per Source Z)" and the reader can verify in seconds. A separate dossier breaks the citation-to-claim adjacency and creates inventory rot.
Citation format: prefer named author + year + title + (chapter/section) for books and reports; URL for vendor docs; DOI or stable URL for academic papers. The drain7 proposal §4 is a worked example of a 5-citation block.
Principle 3 — The novelty test for external citation
Ask: "Would a future reader, reading this canonical/rule/skill cold, encounter a position they could not derive from prior Holistika doctrine + first-principles thinking?"
- Yes → novel framing → external citation MANDATORY (3-5 cites for the framing).
- No → refinement of prior Holistika ratification → external citation OPTIONAL (the prior decision-register row IS the precedent).
Refinements include: v2 extensions of v1 doctrine, tightenings of existing rules, new specialty mints that follow the established quartet pattern. Novel framings include: new pillars (e.g., Quality Fabric architecture), new mechanical contracts (e.g., the rule × skill pairing contract), new posture defaults (e.g., Option-5 conflict surfacing).
Principle 4 — Wave-closure research enrichment is structural, not ornamental
Every wave-close UAT (per akos-planning-traceability.mdc §"UAT quality bar" §10 + parent rule RULE 3) carries a ## Research enrichment subsection naming:
- Which canonicals were enriched with new research material during the wave.
- Which canonicals were identified as needing enrichment but deferred to a future wave.
- Which external sources were newly surfaced during the wave that should inform future authoring.
The subsection is brief (5-15 lines) and makes the operator's lived research practice trackable across waves.
Per-canonical-class craft
Canonical CSV mint (RULE 1)
Authoring sequence:
- Internal evidence-sweep: Grep + Read every existing canonical CSV + every decision row + every process_list row + every baseline_organisation row that bears on the new register. Document the inventory in the ratifying decision row's
rationale cell or in a planning-notes file.
- Novelty test: is the new register a clone-of-existing-pattern (e.g., new sibling adapter registry following Normalised Adapter Pattern) or a novel framing? Clone → optional external cite. Novel → mandatory 3+ external cites.
- External research-sweep: if novel,
WebSearch 5-10 queries on the new register's domain. Capture authoritative sources. Cite at least 3 in the canonical's Evidence-base section.
- Inline citations in
linked_canonicals: frontmatter for internal precedents.
- Evidence-base section in the canonical body for external citations + the novelty rationale.
Doctrine canonical / cursor rule / skill mint (RULE 2)
Same as CSV mint above, but the novelty test is the gating mechanism. Refinements need 0 external cites (Holistika precedent suffices); novel framings need 3-5 external cites.
The drain7 proposal itself is a worked example: rule × skill pairing as a workspace-binding contract is novel enough to warrant 5 citations (ACT-R + Pattern Language + SRE Book + SECI + Anthropic Skills); the per-skill mints inside drain7 are refinements (each skill is a refinement of inline-ratify-craft's shape) so no per-skill external cites.
Wave-closure research enrichment subsection authoring
Template:
## Research enrichment (Wave <letter>)
### Canonicals enriched this wave
- `<canonical-path>` — added §"Evidence base" rows for [topic]; cites [Source A] + [Source B] (per RULE 1 backfill protocol).
### Canonicals identified as needing enrichment (deferred)
- `<canonical-path>` — forward-charter to Wave <letter+1> via OPS-NN-MM (next-review-trigger: <condition>).
### External sources newly surfaced this wave
- [Source A: full citation + URL] — relevant for [domain X canonicals]; consider citing in future mints.
- [Source B: full citation + URL] — relevant for [domain Y canonicals].
Keep this subsection ≤ 15 lines per wave-close.
Pre-flight checklist (before any canonical/rule/skill mint)
- Internal evidence-sweep completed (
Grep + Read + Glob over relevant repo state).
- Novelty test answered (yes → external cites mandatory; no → optional).
- If novel: external research-sweep completed (5-10 queries; sources captured).
- Inline citations integrated into canonical body (not separate dossier).
linked_canonicals: frontmatter populated for internal precedents.- Evidence-base section (or equivalent) carries external citations.
- Wave-closure UAT will include
## Research enrichment subsection.
- Operator-discretion exemption checked (commit message carries
rule-exempt-research-discipline: <reason> only when time-pressed AND backfill-deferred-and-tracked).
Anti-patterns (forbidden by parent canonical §7)
- AP1 — Research-as-ornament: citing 1 paper to look rigorous, without integrating the source's actual position into the doctrine. The cite is decorative; the doctrine ignores the source.
- AP2 — Research-after-the-fact-to-justify: drafting doctrine first, then googling cites that match. Inverts the discipline; the doctrine is unfalsifiable.
- AP3 — Research-as-procrastination: spending 5 hours on WebSearch instead of authoring; the research becomes the deliverable, the doctrine never lands.
- AP4 — Copy-paste-citations: citing a source's abstract without reading the source. The cite is unverifiable; future readers can't validate.
Recovery patterns
- R1 — Time-pressed mint: invoke operator-discretion exemption via
rule-exempt-research-discipline: <one-clause reason> in commit message; backfill in next wave's research enrichment subsection.
- R2 — Research-sweep returns nothing useful: collapse to "the field is silent on this" + cite the search query + cite at least 1 first-principles reasoning chain. Acceptable when the doctrine is genuinely first-of-its-kind.
- R3 — Found contradicting sources: surface as inline-ratify gate (per
akos-inline-ratification.mdc) — let operator choose framing rather than silently pick one.
Cross-references
Source: FraysaXII/openclaw-akos — distributed by TomeVault.
