Skip to content

feat(skills): add aidd-genesplice evolutionary optimization skill#181

Draft
ericelliott wants to merge 6 commits intomainfrom
cursor/new-genesplice-skill-4389
Draft

feat(skills): add aidd-genesplice evolutionary optimization skill#181
ericelliott wants to merge 6 commits intomainfrom
cursor/new-genesplice-skill-4389

Conversation

@ericelliott
Copy link
Copy Markdown
Collaborator

@ericelliott ericelliott commented Apr 7, 2026

Summary

Adds the /genesplice skill — an evolutionary optimization framework for creative output. Closes #180.

File Structure

ai/skills/aidd-genesplice/
├── SKILL.md
├── README.md
└── references/
    ├── candidate-output.md
    ├── quality-gate.md
    └── learnings.md

What's in Each File

SKILL.md — Frontmatter, Competencies block, 4-step SudoLang process (researchCriteria → generationLoop → scoreAllCandidates → suggestWinner), Constraints, and Commands (/genesplice, /genesplice score, /genesplice splice).

README.md — What/why/commands/when-to-use. Includes a "Why" section explaining the genetic algorithm rationale.

references/candidate-output.md — Prototypes folder structure, {short_id} format, SCORES.md tracking convention.

references/quality-gate.md — A11y scorer flow for UI prototype candidates, grade→score mapping table, weighted ×2 rule.

references/learnings.md — Generation size sweet spot, why mutations win, always-preview rule, and the Emergent Redundancy Problem with mandatory holistic criteria.

Validation

frontmatterTokens : 77   (threshold: 100)
bodyLines         : 97   (guideline: 160, hard limit: 500)
bodyTokens        : 867  (guideline: 5000)
✔  Validation passed.

All 462 unit tests pass.

Open in Web Open in Cursor 

cursoragent and others added 6 commits April 7, 2026 23:13
Adds the /genesplice skill — an evolutionary optimization framework for
creative output. Closes #180.

- SKILL.md: frontmatter, constraints, quality gate, commands
- references/algorithm.md: full pipeline, candidate output spec, scoring
  template, learnings including emergent redundancy problem
- README.md: what/why/usage/when-to-use

Validation: 74 frontmatter tokens, 52 body lines, 505 body tokens — well
within all thresholds.

Co-authored-by: Eric Elliott <support@paralleldrive.com>
- Split algorithm.md into candidate-output.md, quality-gate.md, learnings.md
- Rewrite SKILL.md with SudoLang step functions and Competencies block
- Expand README with Why section and full command reference
- Add /genesplice score and /genesplice splice as first-class commands

Validation: 77 frontmatter tokens, 97 body lines, 867 body tokens.

Co-authored-by: Eric Elliott <support@paralleldrive.com>
Updated candidate output documentation for clarity and structure.
…n guidance

- Extract /genesplice research as a standalone subcommand: f(context) => reviewedCriteria
- Require web search favoring research papers, peer-reviewed studies, and
  authoritative industry findings over opinion pieces
- Each criterion outputs one paragraph of explanation + citations (author, title, URL/DOI)
- Reject any criterion that cannot be cited
- Update process pipeline: researchCriteria => research
- Add /genesplice research to Commands block

Co-authored-by: Eric Elliott <support@paralleldrive.com>
…ory review pause

- n parameter = number of generations (not candidates-per-generation)
  2 candidates/generation is the fixed sweet spot per learnings.md
- Add optional userCriteria? to genesplice() and research() signatures
  User-supplied criteria are locked in first and are not overridable by research
  Research fills gaps not covered by userCriteria
- Skill runs autonomously by default; no mandatory user review steps
- Update command signature: /genesplice [-n=2] [context]

Co-authored-by: Eric Elliott <support@paralleldrive.com>
…t stage

Pipeline: gatherContext => research => runGenerations(n) => summarize

- Add gatherContext (effect): scan workspace for existing artifacts, extract
  user-supplied criteria from freeform context prompt
- Split research into fetchBestPractices (effect) + synthesizeCriteria (thinking)
  fetchBestPractices skips web search if enrichedContext already has criteria
- Split generationLoop into buildCandidates (effect) + scoreAndSplice (thinking)
- Criteria now come from freeform context; userCriteria? param removed
- Add 'Run autonomously by default' to Constraints
- Tag each stage with effect/thinking so eval tests can target them correctly

Co-authored-by: Eric Elliott <support@paralleldrive.com>
cursor Bot pushed a commit that referenced this pull request Apr 10, 2026
Tracks #182 — creates tasks/aidd-critical-dependencies-epic.md listing
the dependency chain (PR #168#181#184) and the order of operations
for landing /aidd-parallel, /aidd-genesplice base, and genesplice epic.

Co-authored-by: Eric Elliott <support@paralleldrive.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Add /genesplice to AIDD Framework

2 participants