Skip to content

Arkheionx v5.0.0 — Blind Spot Intelligence

Choose a tag to compare

View all tags
@Yudis-bit Yudis-bit released this 08 Jun 00:11
· 46 commits to main since this release
v5.0.0
6254bda

Arkheionx v5.0.0 — Blind Spot Intelligence

Status: release metadata finalized locally; pending founder push, tag, GitHub
release, and site deploy. The package version is 5.0.0. The last published
tag remains v4.0.0, and the source installers and the GitHub Action still pin
to the last actually-tagged stable release, v3.1.0. Nothing here is published
to PyPI.

1. What V5 adds

V5 adds a Blind Spot Intelligence layer on top of the stable v4 review-map
workflow and the v4.1 research-memory commands. It is the first major step beyond
mapping:

  • V4 maps the protocol — where value moves, the assumptions that guard each
    path, and which paths have no tests.
  • V5 maps where research attention is weakest relative to how much could go
    wrong if a bug existed there.

2. Why V5 exists

A reviewer's scarcest resource is attention. V5 is an attention-allocation engine:
it ranks high-impact surfaces with weak review evidence, estimates criticality
potential (blast radius), and turns guarding assumptions into testable
counterfactuals, so a human or an AI agent spends limited time where it matters.

3. New commands

  • arkheionx blind-spots — likely blind-spot candidates with a transparent
    additive score (impact + review-gap + complexity + assumption).
  • arkheionx criticality-map — criticality potential (heuristic blast radius,
    not severity) across surfaces.
  • arkheionx counterfactuals — "what if this assumption is false?" research
    prompts with local test directions and stop conditions.
  • arkheionx research-pack — a complete local, vendor-agnostic research pack
    (headline). Writes by default.

4. Compatibility

review-map and its focused views (value-paths, assumptions, test-gap-map,
proof-plan) and the v4.1 research-memory commands (agent-brief,
hypothesis-log, case-study) are unchanged. The v5 commands build directly on
review-map output.

5. Quickstart

python3 -m pip install -e .
arkheionx review-map      examples/blind-spot-fixture
arkheionx blind-spots     examples/blind-spot-fixture
arkheionx criticality-map examples/blind-spot-fixture
arkheionx counterfactuals examples/blind-spot-fixture
arkheionx research-pack   examples/blind-spot-fixture --out .arkheionx/research-pack

6. V5 demo

examples/blind-spot-fixture is a small generic protocol — PriceOracle,
CreditVault, ClaimGate, BundleRouter, MockToken — whose test covers only
deposit. The value exit, the oracle/debt path, liquidation, the signature and
Merkle authorization surfaces, and the periphery batch are deliberately untested,
so ArkheionX surfaces them as likely blind spots. No planted vulnerabilities.

7. JSON schemas

schemas/blind-spots.schema.json, schemas/criticality-map.schema.json,
schemas/counterfactuals.schema.json, and
schemas/research-pack-manifest.schema.json. Every JSON output is parseable and
schema-validated by the test suite.

8. Safety boundaries

Local and static only. No RPC, no live-chain calls, no exploit automation, no
private-key handling, no auto-submit. Blind spot candidates are not
vulnerabilities; criticality potential is not severity; counterfactuals are
research prompts, not findings. ArkheionX does not confirm vulnerabilities,
assign final severity, prove protocol safety, or replace an audit. Human review
is required.

9. Known limitations

  • Heuristic and static; blind spot detection is not perfect.
  • No bug prediction, no severity, no probability.
  • Real-protocol validation is still required
    (../REAL_PROTOCOL_PROOF_PLAN.md).
  • A finding is only ever confirmed by a human with independent local proof.

10. Validation matrix

Docs links, safety wording, version consistency, release readiness, the full
unit-test suite, make validate, the package build (python -m build +
twine check), and the website build all pass.

11. Documentation

12. GitHub release draft

Arkheionx v5.0.0 — Blind Spot Intelligence. V5 adds a local/static layer
that prioritizes high-impact surfaces with weak review evidence, generates
counterfactual research prompts, and packages the result into an AI/human-ready
research pack. It builds on the stable review-map workflow. No RPC, no
exploit automation, no severity, no bug claims; human review required.

13. Founder release commands

The package version is already finalized at 5.0.0 in this tree. To release:

git status --short --branch
make validate
git push origin main
git tag -a v5.0.0 -m "Arkheionx v5.0.0 — Blind Spot Intelligence"
git push origin v5.0.0
gh release create v5.0.0 \
  --title "Arkheionx v5.0.0 — Blind Spot Intelligence" \
  --notes-file docs/releases/V5_RELEASE_NOTES.md \
  dist/arkheionx-5.0.0-py3-none-any.whl \
  dist/arkheionx-5.0.0.tar.gz

The last tagged stable release stays v3.1.0 (the installer/action pin) until a
new stable tag is cut.

Assets 4
Loading