Write GenesisCode Skill Distribution Kit v1

Executable multi-agent distribution kit for GenesisCode authoring.

This package is the runnable companion to:

• docs/spec/WRITE_GENESISCODE_SKILL_PACK_v0.1.md
• docs/spec/WRITE_GENESISCODE_SKILL_v0.1.md

Canonical Kit Root

• docs/skill_pack/write_genesiscode_v1/manifest.json

Kit Contents

• Hand-owned workflow source: policies/genesiscode_authoring_workflow_v0.1.json
• Generated compact index: docs/skill_pack/write_genesiscode_v1/authoring-card.md
• Generated prompt registry: docs/skill_pack/write_genesiscode_v1/prompt-cards.json
• Generated executable recipe registry: docs/skill_pack/write_genesiscode_v1/recipe-cards.json
• Generated provenance manifest: docs/skill_pack/write_genesiscode_v1/manifest.json
• Deterministic renderer/checker: scripts/lib/genesiscode_authoring_skill.py
• Transactional update entrypoint: bash scripts/update_agent_authoring_bundle.sh authoring-skill

The registries replace per-card prose files. Their source digest and every canonical input identity are bound in the manifest, and stale output fails closed. - Deterministic verification entrypoint: - scripts/check_write_genesiscode_skill_distribution.sh

Work Routing

The installed skill must preserve the closed user-task, active-defect, roadmap-task, and exploratory-work route set. In particular, an empty upgrade_plan.md disables the defect route; it never redirects an agent to wait, invent defects, or treat status views as task authority. General roadmap work is selected from the machine-generated execution slice, while exploration carries no completion authority.

Runtime Verification Contract

When GENESIS_WRITE_SKILL_DIST_VERIFY_RUNTIME=1, the verification script must enforce:

• genesis/write-genesiscode-skill-conformance-v0.1 report kind.
• Minimum conformance score from kit manifest distribution requirements (>= 100 by default).
• Minimum corpus breadth thresholds from manifest (min_prompts, min_recipes).
• Required domain coverage from manifest (required_recipe_domains) including:
  • service
  • game-loop/graphics
  • gpu-compute
  • gpu non-graphics compute
  • package publish/sync
  • deployment targets
  • failure recovery
  • performance triage
  • assurance
  • plugin/ffi
  • xr runtime
  • xr productization/deploy-test
  • durable data
• At least one fault-injection recipe (mode = "fault-injection").

Integration

This v1 kit is a primary AI entrypoint and must stay listed in:

• docs/spec/AGENT_AUTHORING_BUNDLE_v0.1.md
• docs/AGENT_ONBOARDING_v0.1.md