nw-ab-house-style
KNOWLEDGE — caveman-native authoring house style + by-construction guarantees (Reasoning Mandate injection, measured-gain compression). Reference for create/migrate; no sequence.
SKILL.mdnWave/skills/nw-ab-house-style/SKILL.md
GitHubnw-ab-house-style (KNOWLEDGE)
Kind: KNOWLEDGE (reference). Loaded by nw-ab-create-agent (CREATE) and nw-ab-migrate-monolith (EXTRACT). No forced sequence.
House style
Author caveman-curated: dry, declarative, tables and compact one-line bold-lead lists, zero narrative padding. Agent body lean (~100-200 lines, role+routing+contract); deep knowledge in skills (<5000 tokens each). Exemplar: nw-security-analyst — 104-line agent (93 body), deep knowledge in 2 skills (96+107=203). Byte-exact preservation: ### Example N: headers, NORMATIVE blocks, code/YAML, AskUserQuestion trees, machine content. User-facing templates stay clear, not compressed.
By-construction guarantees
- House style — every created/modified asset is caveman-curated (above).
- Reasoning-mandate injection — every created/modified agent gets the
## Reasoning Mandateblock inserted verbatim. Depth from rigor, never padding. - Opportunistic retro-compression (measured-gain ≥20% or skip) — on MODIFY, measure prose share and compress only for meaningful reduction while preserving deliberate public/machine contracts. Preserve
### Example N:and normative machine content where consumed. Incidental wording/hash/content-pin tests are not authority: remove them rather than repin changed prose. Verify reachable role/skill references, constructor/generated contract consistency and actual rendered/installed consumer behavior; never bump a hash to sell semantic assurance.
Reasoning Mandate block (inject verbatim)
## Reasoning Mandate (Caveman)
Verdict-first, tables over prose, evidence-dense, zero narrative. Depth comes from rigor, not padding. State the conclusion, then the supporting evidence; never bury the verdict under exposition.