Skip to content

v0.1.0 -- initial public release

Latest
Latest

Choose a tag to compare

View all tags
@autogame-17 autogame-17 released this 11 May 07:05
· 1 commit to main since this release
v0.1.0
85fcf24

Initial public release of skill2gep -- a protocol adapter that turns a locally-executed Skill (Cursor skill, Claude / Anthropic skill, or any procedural SKILL.md) into GEP (Genome Evolution Protocol) assets:

  • Gene -- a compact strategy template (signals_match + strategy + AVOID: + validation)
  • Capsule -- an auditable record of one real execution of that Gene (outcome + execution_trace + env_fingerprint)

Skills are written for humans; GEP assets are written for model execution. skill2gep packages both sides and refuses to fabricate a Capsule from the document alone -- the Capsule only exists when a real execution backs it.

Based on Wang, Ren, Zhang, From Procedural Skills to Strategy Genes: Towards Experience-Driven Test-Time Evolution (Infinite Evolution Lab / EvoMap x Tsinghua University), arXiv:2604.15097.

What is in this release

  • SKILL.md -- the full 10-phase distillation workflow (read Skill, dedup, draft candidate Genes, hard-gate validate, install, run, collect Capsule, optionally publish)
  • scripts/validate_gene.js -- zero-dependency Gene validator (schema, AVOID requirement, token ceiling, private-path leak guard)
  • scripts/validate_capsule.js -- zero-dependency Capsule validator (forgery guard, validation-coverage check, dangling-reference check)
  • examples/ -- worked examples
  • test/ -- regression scenarios for the validators

Honest scope note

  • skill2gep is a protocol adapter, not a magic distiller. Phases 1-8 are still executed by the calling agent against the source SKILL.md text; the tool formats the result into GEP schema, runs hard-gate validators, and ships it.
  • arXiv:2604.15097 validates Gene-as-control-interface on 45 scientific code-solving scenarios with Gemini 3.1 Pro and Flash Lite. Claims about other domains (web automation, long tool chains, multi-agent negotiation) are extrapolation and are flagged as _source.claims_outside_scope: "assumption" in the emitted Gene.
  • "Gene format outperforms Skill format on the paper's task set" is not the same as "skill2gep reliably produces high-quality Genes on your workload." When you publish to EvoMap, community validators score the asset; low-quality Genes get rejected or down-weighted.

evolver CLI integration

If you have @evomap/evolver installed, you can run the reverse-distillation pipeline directly, with sandboxed storage and forgery guards, without involving an agent loop:

# Gene-only, no publish
evolver skill2gep ./path/to/skill --no-publish

# With real execution evidence -> Gene + Capsule
evolver skill2gep ./path/to/skill \
  --execution=./skill-run.json \
  --platform=cursor

# Strict mode: refuse Skills whose validation is not node/npm/npx
evolver skill2gep ./path/to/skill --strict

See SKILL.md for the execution JSON shape and the full 10-phase workflow.

Install

As a Cursor Skill

git clone https://github.com/EvoMap/skill2gep.git ~/.cursor/skills/skill2gep

As an npm package

npm install -g @evomap/skill2gep

Exposes the two validator binaries skill2gep-validate-gene and skill2gep-validate-capsule.

License

MIT. Designed as a protocol adapter to encourage integration (including in closed-source agents). The EvoMap runtime layer (evolver, evomap-hub, gep-sdk-js) is GPL-3.0-or-later.

References

Assets 2
Loading