Add shared glossary for AIEPFD plugin - #1975
Conversation
f7a980d to
c5b638d
Compare
| interpreting or producing plugin artifacts so terms are used consistently across | ||
| skills. | ||
|
|
||
| This glossary mirrors the terminology in the Confluence page _Glossario del |
There was a problem hiding this comment.
thoughts: as every other template is in english we may leave even this glossary in english
51f47c4 to
8a6c43e
Compare
The prd-alchemist skill takes unstructured product input (chat logs, meeting notes, documents, scattered ideas) and synthesizes it into a coherent first PRD draft. It gathers material, analyzes sources to find shared meaning, regroups related information, surfaces gaps explicitly, and produces a Status:draft document. The skill is the gathering phase only: it does not publish, review, or advance the document beyond draft status. Includes: - SKILL.md: gathering workflow with guardrails - templates/prd-draft.md: lightweight draft scaffold - plugin.json: aiepfd plugin with skills registration Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Address PR review feedback: - Expand the acronym to "Product Requirements Document (PRD)" in the skill description and add synonyms (product spec, requirements document) so the skill activates on more natural user phrasings. - Add an "About PRDs" section giving the agent framing on what a PRD is, its purpose, and its audience, so drafts match the tone and focus expected by product managers, engineers, and designers. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Add full metadata frontmatter and stable machine-readable IDs (YAML keys plus per-section id anchors) so extraction agents can read and update specific fields without parsing free text. Add a Linked artifacts section and split open questions into resolved and still-open subsections. Support two entry modes (material-first and interview-first) and require clarifying questions whenever input is unclear, never inventing content. Treat "draft" only as a metadata status value rather than a document type, so a later stage can advance the PRD; rename the template to prd.md. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Keep plugin terminology in one place so skills can reuse a single glossary instead of duplicating definitions.\n\nThis wires the prd-alchemist skill to read the shared glossary before proceeding, which makes term usage more consistent across future plugin skills. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Keep the plugin glossary consistent with the shared framework terminology so skills can rely on the same definitions used in Confluence.\n\nThis replaces the initial lightweight glossary with the framework's acronym, concept, and role taxonomy. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
25ee412 to
1b09899
Compare
Keep shared terminology aligned with the English plugin templates. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Attach the shared glossary to the existing aiepdf plugin instead of creating a duplicate aiepfd tree. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Ensure domain-authoring skills use the same framework terminology. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
lucacavallaro
left a comment
There was a problem hiding this comment.
@kin0992, could you please explain the rationale behind this change? The glossary has ~2,000 tokens, which will be loaded upfront with every skill invocation.
I noticed that some terms in the glossary are very common (e.g., CI, CI/CD), so we might not need to instruct the model on these concepts.
Have you conducted any evals against these skills to measure the performance impact?
@lucacavallaro the goal of this PR is to provide a common glossary to the skills of the plugin, so that is it not necessary to duplicate it inside them. |
Add a plugin-level glossary for
plugins/aiepfdand have theprd-alchemistskill read it before proceeding.