Implement cellix-tdd agent skill

[nnoce14]

Objective

Create a new agent skill under .agents/skills/cellix-tdd/ for consumer-first, TDD-driven development of @cellix/* framework packages.

Deliverables

• .agents/skills/cellix-tdd/SKILL.md
• .agents/skills/cellix-tdd/rubric.md
• .agents/skills/cellix-tdd/fixtures/...
• .agents/skills/cellix-tdd/evaluator/evaluate-cellix-tdd.ts
• any necessary repo docs/index updates so the skill is discoverable

Required behavior

The skill must enforce this workflow:

consumer usage discovery → package intent alignment → public contract definition → failing tests against public APIs only → implementation/refactor → documentation alignment → release hardening → validation

TDD is central to this skill:

• expected behavior should be clarified before implementation
• public-contract tests should be written before implementation where behavior is being added or changed
• implementation should emerge from the tests and expected consumer usage
• refactors should preserve contract tests unless the public contract is intentionally changed

Discovery expectations

During the initial discovery phase, the skill should inspect existing package/repo context first.

If expected behavior, consumer usage, or package boundaries are materially unclear, the skill should collaborate with the user up front to clarify the contract before authoring tests and implementation.

That clarified understanding should be captured in manifest.md and then drive the test plan.

Non-negotiables

• target packages must maintain manifest.md
• README.md must remain consumer-facing
• meaningful public exports must have useful TSDoc
• TSDoc should document description/purpose, signatures or type intent, and expected usage as necessary, plus params/returns/errors/examples where relevant
• tests must verify behavior through public APIs only
• public behavior/export changes require docs alignment
• the skill must leave a validation summary

Harness requirements

The evaluator must score observable outputs/artifacts and verify:

• no internal/deep-import testing
• manifest/README/TSDoc alignment
• required workflow sections in skill output
• release-hardening notes
• rubric-based pass/fail reporting

Minimum fixture coverage

• existing package add feature
• existing package internal refactor
• new package greenfield
• docs lagging implementation
• leaky/overbroad API
• tempting internal helper

Acceptance criteria

• cellix-tdd exists and is discoverable
• workflow and rules are encoded in SKILL.md
• evaluator and fixtures exist
• rubric-based reporting works
• documentation requirements are enforced
• public-contract-only testing is enforced

See for additional task context:
Note: These will be removed after this task is completed; relevant details must be captured in the cellix-tdd skill

• .agents/docs/cellix-tdd/context.md
• .agents/docs/cellix-tdd/package-docs-model.md
• .agents/docs/cellix-tdd/package-manifest-template.md