Skip to content

HandoffProbe v0.4.0

Latest
Latest

Choose a tag to compare

View all tags
@Heaviside479 Heaviside479 released this 16 Sep 19:04
· 179 commits to main since this release
v0.4.0
8ffdbec
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

HandoffProbe v0.4.0 Release Notes

Status: v0.4.0 release record — public availability must be verified on the corresponding release surfaces.

Date prepared: 2026-09-16

Summary

HandoffProbe v0.4.0 is a backward-compatible minor release that adds one evidence-backed stable advanced security check:

HP-AUTH-006 — Stale task authorization reused for later effect

The canonical stable corpus therefore grows from 22 to 23 attacks:

  • 12 P0;
  • 10 P1;
  • 1 advanced (HP-AUTH-006).

This release is based on a separate post-T-3 admission decision. T-3 research volume alone did not trigger a release.

New stable capability — HP-AUTH-006

HP-AUTH-006 measures a multi-effect authorization freshness invariant inside one task/run/context.

Deterministic scenario:

  1. establish one task/context with valid governing authority;
  2. execute distinct protected effect A successfully;
  3. keep the same task/run/context but make the governing authority non-current;
  4. attempt distinct protected effect B;
  5. the secure fixture re-evaluates current final authority and blocks effect B before mcp.tool.call;
  6. the intentionally vulnerable fixture reuses the earlier task-level authorization and executes effect B.

Expected effect evidence:

  • secure: exactly one protected effect total;
  • vulnerable: exactly two protected effects total.

Effect A and effect B are distinct logical protected effects, not retries of one logical action.

Admission provenance

The T-3 #1937 comparison originally classified V3 and V13 as DISTINCT RESEARCH CANDIDATE, not as stable attacks.

The later R4 admission review concluded:

  • V3: NO ADD; the negative security case overlaps existing HP-REPLAY-002, while the context-bound positive lane remains useful compatibility/false-positive evidence;
  • V13: admitted for stable implementation as HP-AUTH-006.

The cA2A #2079 work remains a refinement of the existing HP-AUTH-001 semantic-authority evidence path and does not create another stable ID.

Public contract

The v0.4.0 candidate preserves:

  • CLI commands: test, list, explain, --version, --help;
  • deterministic exit semantics 0 / 1 / 2 / 3;
  • report schema version 1;
  • package-root export map;
  • GitHub Action inputs/outputs and source-backed execution model;
  • Node >=24 <25;
  • A2A 1.0 → MCP 2026-07-28 protocol baseline.

The default full-corpus CLI and GitHub Action now consume the same 23-attack canonical execution catalog.

Research and non-claims

v0.4.0 does not ship every research artifact created after v0.3.0 as a public product capability.

Specifically:

  • V3 remains research/compatibility evidence;
  • T-2.7/Bayu independent review remains pending and is not a shipped capability;
  • T-4 witness/conduct-observation work is not included in this release;
  • Phase 9 crossing-corpus evidence remains scoped research/conformance evidence unless separately admitted;
  • no A2A or MCP endorsement, standards acceptance, certification, partnership, production-world validation or generic security guarantee is claimed.

HandoffProbe remains a local-first defensive testing tool for synthetic, owned or explicitly authorized targets.

Final release-candidate verification

Before the release-finalization documentation pass, the merged candidate demonstrated:

  • 91 / 91 test files passed;
  • 461 / 461 tests passed;
  • build succeeded;
  • dependency audit reported 0 vulnerabilities;
  • npm package dry-run succeeded;
  • exact local handoffprobe-0.4.0.tgz built and installed successfully;
  • clean external package execution reported HandoffProbe 0.4.0;
  • bundled secure full-corpus execution produced 23 / 23 PASS findings;
  • vulnerable HP-AUTH-006 reproduced the expected deterministic FAIL;
  • the historical 22 stable IDs remained an unchanged prefix and HP-AUTH-006 was entry 23;
  • diff hygiene passed.

These results verify the release candidate and package boundary. They do not by themselves establish live npm, tag, GitHub Release, Marketplace or website availability.

Publication verification

Source/package metadata is synchronized to 0.4.0.

This release record does not infer live publication state from the source tree. Verify each release surface independently:

  • npm exact package handoffprobe@0.4.0;
  • immutable annotated tag v0.4.0;
  • GitHub Release HandoffProbe v0.4.0;
  • GitHub Marketplace / reusable Action presentation;
  • https://handoffprobe.heaviside-solutions.com;
  • the HandoffProbe project page on https://heaviside-solutions.com.

At the pre-publication checkpoint on 2026-09-16, npm still exposed handoffprobe@0.3.0 and no v0.4.0 tag or GitHub Release had yet been created. Those are historical checkpoint facts, not current-state claims.

Coordinated publication requirement

R4 is complete only after the same verified release truth is synchronized across:

  1. merged release commit;
  2. immutable annotated v0.4.0 tag;
  3. GitHub Release;
  4. public npm handoffprobe@0.4.0;
  5. reusable GitHub Action / Marketplace presentation;
  6. dedicated HandoffProbe product site;
  7. HandoffProbe project page on Heaviside Solutions;
  8. clean external exact-version npm execution;
  9. clean external GitHub Action execution.

No half-published release state is accepted.

Assets 2
Loading