training-report

Training-report produces a professional training or workshop report delivered as a .docx file. It is intended for any request to document a training session, workshop, or onboarding event with individual participant feedback and recommendations, and it triggers on phrases like "training report", "workshop report", "compte rendu de formation", or "I just ran a workshop and need to write it up". The skill is discipline-agnostic, applying equally to coding workshops, leadership seminars, safety training, creative sessions, and onboarding, for both short and long sessions.

The core principle is to iterate the full report in Markdown first and generate the .docx only once at the end, treating the Markdown as the canonical artifact and the Word file as a terminal derivative. Before asking the user anything, it checks dependencies: the docx skill is required for the final document, and a language-matched humanizer skill is recommended, with inline humanization rules as a fallback. It then confirms the report language and primary reader, loads a tone-of-voice reference, and asks whether the user has a branded .docx template to use as the base, defaulting to a clean document with a configurable brand color.

Most of the work is a structured interview conducted in batches, waiting for answers between them and extracting what the user has already said before asking. Batches cover session metadata, context and goals, starting skill levels, a step-by-step session walkthrough, deliverables produced, optional general observations, optional individual participant feedback framed diplomatically, recommendations and next steps, annexes, and an optional closing note with contact details. Annex handling distinguishes embeddable images from referenced files and dedicated survey sections. A voice-mode note reminds the author to clarify ambiguous transcribed names or terms rather than guess. Reference files for tone of voice, the Markdown draft, and docx generation are loaded only at the steps where they apply. The skill is MIT licensed and user-invocable.