Revert to 0.14.0, remove internal strategy docs, honest README status, fix ADK tools fail-open

[pengfei-threemoonslab]

Summary

Reconcile the version story and tighten the pre-1.0 posture. Four related changes:

• Version 1.0.0a1 → 0.14.0, made consistent everywhere. The engine is deterministic and well-built, but a 1.0 label was indefensible: the report contract is still report_schema_version: "0.28", the classifier is still 4 - Beta, the last released tag is v0.13.0, and no real-world detection-accuracy baseline is published. Reverted pyproject/__init__ and the ~84 files that pin or stamp the version (install snippets, .well-known, schema defaults, 17 version-stamped goldens, 4 test assertions, regenerated llms-full.txt). tool.version, cli_version, and the generated Action pin all derive from __version__, so they moved in lockstep. STABILITY.md's migration note is rewritten as a 0.x continuation (not a new "alpha" line); CHANGELOG Unreleased becomes a dated 0.14.0 section documenting why the 1.0 line was withdrawn.
• Removed internal strategy / GTM material from the public repo (59 files): docs/shipgate-strategic-engineering-review.md (internal SWOT), docs/decks/ (incl. the VC deck PDF/PPTX), and marketing/. Referrers reconciled (CONTRIBUTING, pyproject sdist + ruff excludes, .gitignore). Note: removed from HEAD only — git history still contains them.
• Honest README framing. Added a pre-1.0/beta status banner and a "…and here's the failure mode" note after the two-PR demo: real dynamic tool surfaces return insufficient_evidence by design, and detection accuracy is still being validated.
• Fixed a fail-open in inputs/google_adk.py. A tools: value that is present but not a list (templated string, env var, mapping) in an Agent Config silently collapsed to a confident tool_count: 0. It now records a warning + a dynamic GoogleAdkToolset marker (mirroring the Python entrypoint), routing to SHIP-ADK-DYNAMIC-TOOLSET-NOT-ENUMERABLE instead of a silent zero. Regression test added. Audited every other input parser for the same else [] shape — the rest are benign sub-field coercions.

Type

• Input adapter change
• CLI or GitHub Action behavior
• Documentation only
• Check or risk-model change
• Report, schema, or SARIF output

Verification

CI is authoritative for python -m ruff check ., python -m compileall -q src tests, and python -m pytest.

Additional local checks run:

• ruff check . → clean; compileall of changed source → OK
• Full pytest -n auto -m "not perf" → green (1 skip), incl. version-stamped goldens, the 5 refreshed render-hash snapshots, test_docs_links, test_public_surface_contract, and the new ADK regression test
• shipgate --version → Agents Shipgate 0.14.0

Release-readiness notes

• No user-code import added to default scan paths
• No network access added to default scan paths
• New or changed check IDs are documented in docs/checks.md — no new check IDs; the ADK fix routes to the existing SHIP-ADK-DYNAMIC-TOOLSET-NOT-ENUMERABLE
• Report/schema changes are additive or documented in STABILITY.md — no report-schema change; STABILITY.md header + migration note updated for the 0.14.0 line