New enhance-prompt workflow: continuously enhance existing prompt assets with current PromptKit capabilities

[abeltrano]

Problem / Goal

PromptKit can generate LLM prompt assets (raw prompts, agent instruction
files, Copilot prompt files, agentic workflows) from scratch, but there is
no supported way to take an existing prompt asset — a prompt, agent file,
instruction file, or a combination — and enhance or "enlighten" it with
PromptKit smarts.

The input asset may or may not have been generated with PromptKit originally:

• If it was not PromptKit-generated, the goal is to enrich it with
  relevant PromptKit capabilities (personas, guardrails, reasoning protocols,
  output-format discipline, determinism hardening).
• If it was PromptKit-generated, the goal is to update it to
  incorporate PromptKit features and functionality that were not available
  when it was first generated.

The broader objective is to enable continuous improvement of LLM
prompt assets / tooling, so they don't stagnate at the capability level of
whenever they were first authored.

Proposed approach

A new interactive workflow — enhance-prompt — that:

1. Ingests an existing prompt asset (or set of assets) and classifies its type.
2. Detects PromptKit provenance/version if present, and degrades gracefully
   when absent.
3. Inventories the asset's current PromptKit-equivalent capabilities.
4. Performs a gap analysis against the current PromptKit library.
5. Proposes a prioritized, category-classified set of enhancements
   (document mode) and can apply them in place after confirmation
   (action mode) — modeled on review-pull-request's dual-mode pattern.
6. Verifies the result and reports what changed.

The user can optionally scope the work to a specific enhancement area;
otherwise PromptKit analyzes the asset and suggests categories.

Supporting components

Type	Name	Purpose
template	enhance-prompt	Interactive, dual-mode orchestration (category: agent-authoring)
protocol (reasoning)	prompt-enhancement	Ingest → provenance/version detect → baseline → gap analysis → propose → apply → verify
format	enhancement-report	Provenance/version, capability baseline, prioritized enhancement plan, applied changes, residual gaps
taxonomy	enhancement-categories	Classification scheme + user-selectable focus areas

Reuses the promptkit-contributor persona and the standard anti-hallucination
and self-verification guardrails. Complements decompose-prompt (which
extracts library components from a prompt) by going the other direction —
enriching the user's asset from the library.

Low-priority / exploratory follow-ups

These are loose ideas, low priority, to be explored — not committed, and
not prerequisites for the workflow above:

• Internal versioning/tagging of PromptKit-generated outputs, so the
  enhancement workflow can identify which PromptKit features were available
  at generation time vs. now.
• Embedding PromptKit's current version in generated outputs in general
  (e.g., extending the existing <!-- Generated by PromptKit — edit with care --> provenance marker to include a version), which would require
  changes to bootstrap.md and the output-mode formats.

The enhance-prompt workflow must function whether or not a version marker
is present.

Acceptance criteria

• A new template lets a user enhance an existing prompt/agent/instruction
  asset (or a combination) with current PromptKit capabilities.
• Works on both PromptKit-generated and non-PromptKit assets.
• Supports document mode (enhancement plan) and action mode (apply in place
  after explicit confirmation).
• Lets the user optionally scope by enhancement category, or suggests
  categories when none is specified.
• Does not require version-stamping of outputs to function.
• manifest.yaml updated; python tests/validate-manifest.py passes.