feat: add agent-native source-grounded getslide workflow

[1beenlee]

Summary

• reposition getslide.md around a source-grounded deck compiler / Agent Skill workflow while keeping standalone HTML as the canonical artifact
• add project-level getslide skills for Agent Skills/Codex and Claude Code with enforced byte parity
• add zero-dependency arbitrary text/Markdown source staging via tools/prepare-deck.mjs
• add high / medium / low source-sufficiency gates
• prevent changed source from being combined with a stale DECK_BRIEF.md, and remove stale generation packets when a brief is reset
• add post-generation AI editability evaluation probes
• align English/Korean README, AGENTS, product decisions, open-source boundary, and benchmark docs
• add a minimal non-deploying GitHub Actions validation workflow

Scope guardrails

• no changes to examples/, templates/, or prompts/
• no model API, provider routing, runtime network calls, dependencies, auth, payment, queue, analytics, hosted inference, marketplace, or PPTX implementation
• no real-user/private source material
• no deploy or Preview workflow

Validation — PASS

GitHub Actions run 32346730105 on head 5779d45b528cf1b31ab399a827dfc4428ba1a722 completed successfully.

node --check tools/prepare-deck.mjs                       # PASS
node --check tools/test-agent-workflow.mjs                # PASS
node tools/test-agent-workflow.mjs                        # PASS
node tools/test-generation-harness.mjs                    # PASS
node tools/validate-deck.mjs examples/hackathon-demo/index.html  # PASS

The v0.3 regression specifically verified:

• cross-client skill copies are synchronized
• arbitrary source staging produces both workflow packets
• changed source cannot be combined with a stale DECK_BRIEF.md
• removing/resetting the brief removes a stale generation packet
• source staging remains provider-neutral and zero-network

Existing v0.2.1 regression verified canonical navigation protection, generation packet construction, invalid-navigation rejection, the public example deck, and provider-neutral/zero-dependency behavior.

Browser/print visual QA is not a fresh gate for this PR because the existing example/template are intentionally unchanged. docs/EDITABILITY_EVAL.md defines the next real-deck evaluation milestone rather than claiming editability from static validation alone.

Strategy note

Current ecosystem review found that HTML presentation Agent Skills are already an emerging pattern. This PR therefore treats Agent Skill + HTML as distribution/architecture, not the moat. The durable focus is explicit source grounding, visible gaps/assumptions, deterministic contract checks, and safe targeted post-generation edits.