Feat/skills publication

[charlyautomatiza]

This pull request introduces comprehensive support for an "exploratory planning mode" in the k6-plan and k6-builder skills, enhancing their ability to generate provisional performance test plans from partial or ambiguous user input and application context. It also strengthens the policies for handling uncertainty, clarifies the output contract, and ensures that inferred or assumption-based values are never presented as confirmed facts in summaries or runnable artifacts.

The most important changes are:

Exploratory Planning Mode & Workflow (k6-plan):

• Added automatic detection and handling of exploratory mode when user input is incomplete but sufficient context exists to suggest a provisional plan, including new policies for generating, labeling, and outputting exploratory plans (plan_mode: exploratory, ready_for_builder: false, assumption_basis, inferred_targets, inferred_flows, pending_clarifications). [1] [2] [3] [4] [5] [6] [7]
• Updated the main planning workflow to inspect application context for likely test targets and flows when explicit endpoints or SLAs are missing, and to generate prioritized candidate lists for provisional planning. [1] [2]

Output Contract & Guardrails:

• Revised the output contract and validation checklist to require explicit marking and isolation of all inferred or assumption-based values in exploratory mode, and to prevent their promotion to confirmed or runnable status until blockers are resolved. [1] [2] [3] [4] [5]
• Ensured that exploratory plans include all relevant metadata and that summaries clearly distinguish between verified and provisional facts. [1] [2]

Builder Handoff and Safety (k6-builder):

• Updated builder handoff policy to carry forward all exploratory markers and unresolved fields, preventing the generation of executable artifacts from exploratory plans and ensuring that all uncertainties are surfaced as pending questions. [1] [2]

Clarification and Question System:

• Improved the adaptive question system to allow exploratory mode to bypass clarification hard stops when sufficient evidence exists, and to funnel all unresolved critical inputs into a unified pending questions block. [1] [2]

Documentation & Metadata:

• Updated descriptions, repository links, and versioning in plugin.json and skill documentation to reflect the expanded scope and new exploratory features. [1] [2]

These changes make the planning and builder skills more robust, flexible, and transparent when dealing with incomplete or ambiguous requirements, while maintaining strict separation between provisional and confirmed test artifacts.