catalog-v5.0: Merge pull request #25 from drayline/release/v5.0

5.5-generation Claude alignment ship for the rootnode-skills system. 27 Skills recalibrated 2026-10-22.

This release coordinates the v5.0 catalog ship. Each Skill ships its own per-package release; this is the umbrella that frames the catalog as a unified system.

What's in v5.0

27 Skills across 9 functional categories. Two surfaces: chat-side Claude Projects and Claude Code execution. Unified by a single principle layer governing both.

The v5.0 ship is the 5.5-generation alignment cycle triggered by four releases in five weeks: Fable 5.1, Opus 5.5, Sonnet 5.5, and Haiku 5.5. Calibration scope moves to Opus 5.5 + Sonnet 5.5 dual-primary (Opus 5.5 is the Claude Code default on every Anthropic-hosted plan), with Fable 5.1 integrated-aware, Haiku 5.5 subagent-aware, Opus 5 / Sonnet 5 fallback-graceful, and Sonnet 4.6 legacy-graceful. Haiku 4.5 leaves calibration scope; on third-party providers the haiku alias may still resolve to it — pin claude-haiku-5-5 via ANTHROPIC_DEFAULT_HAIKU_MODEL.

Load-bearing changes:

• Effort default moved. Opus 5.5, Sonnet 5.5, and Haiku 5.5 default to medium in Claude Code (Sonnet 5.5 and Fable 5.1 default to high on the Claude API). Anthropic reports Opus 5.5 at medium matches or exceeds Opus 5 at high. Every calibration line is rewritten whole; xhigh/max are reserved for measured gains.
• Thinking can no longer be disabled on Opus 5.5, Sonnet 5.5, or Fable 5.1. Prompt scaffolds drop thinking-disabled mitigations, and instructions that ask the model to reproduce its reasoning in the response are flagged as a reasoning_extraction refusal trigger (prompt-compilation, prompt-validation, reasoning-blocks).
• Tendency #15 — early turn termination. The behavioral taxonomy grows 14 → 15 with the one behavior documented across all four 5.5-generation prompting guides: ending a turn with a summary and no tool call while work is still owed. behavioral-tuning also repairs its routing surface so tendencies 11–15 are reachable from SKILL.md, and carries Anthropic's countermeasure blocks verbatim.
• Hook correctness fix (cc-design, repo-hygiene). cc-design's hook reference taught exit 1 as blocking and a "matchers" schema; enforcement requires exit 2, and config is a matcher string plus a hooks array. repo-hygiene gains the audit-side check that flags exit-1 enforcement as non-blocking.
• Session reports and delegation hardening (cc-design). A computed session-report schema, spec-digest-bound briefs, capped Builder→Refuter loops, and status: blocked as distinct from an empty findings list.
• Per-model fallback semantics. Server-side fallback is now per model — Haiku 5.5 has none; Sonnet 5.5 retries only cyber / frontier_llm declines on Sonnet 5 — and refusal billing changed 2026-09-24.

Every Skill moves to 5.0.0; tier structure (Model-compatible / Sonnet-graceful / High-effort recommended) is unchanged at 13 / 7 / 7. See the Model Compatibility section of the README for effort defaults by model and surface, fallback semantics, and per-tier definitions.

Download

All Skills, one download: grab rootnode-catalog-v5.0.zip from the Assets below. It bundles all 29 packaged Skill files (24 chat-side -cp flat zips for Claude Projects, 5 Claude Code -cc wrapper zips). Unzip it, then upload or install the Skills you want.

Individual Skills: use the links under Skills below — each Skill's own release carries its packaged zip.

GitHub auto-attaches "Source code (zip / tar.gz)" to every release; those are the raw repository, not the packaged Skills. Ignore them and use the bundle or the per-Skill zips.

Skills

Build (3)

• rootnode-prompt-compilation
• rootnode-skill-builder
• rootnode-cc-design

Diagnose (5)

• rootnode-project-audit
• rootnode-global-audit
• rootnode-anti-pattern-detection
• rootnode-prompt-validation
• rootnode-repo-hygiene

Optimize (3)

• rootnode-behavioral-tuning
• rootnode-memory-optimization
• rootnode-context-budget

Cross-Layer (1)

• rootnode-full-stack-audit

Bridge (2)

• rootnode-project-brief
• rootnode-session-handoff

Cross-Boundary (2)

• rootnode-handoff-trigger-check
• rootnode-profile-builder

Runtime (CC, 2)

• rootnode-critic-gate
• rootnode-mode-router

Block Libraries (4)

• rootnode-block-selection
• rootnode-identity-blocks
• rootnode-reasoning-blocks
• rootnode-output-blocks

Domain Packs (5)

• rootnode-domain-business-strategy
• rootnode-domain-software-engineering
• rootnode-domain-content-communications
• rootnode-domain-research-analysis
• rootnode-domain-agentic-context

Architecture

Two surfaces, one mental model. Chat-side Claude Projects and Claude Code execution share a unified principle layer covering placement discipline, decomposition by mechanism, behavioral countermeasures, cross-Skill composition contracts, verification-instruction discipline (D4), and landscape-volatility discipline (D8).

Site

rootnode.design