netstack v0.5.0

v0.5.0

Reviewed commit: 16728fb9d9df7c7486fa85bd173398c167068897.

A deeper view of NetNet: understand how its products work, measure their activity, and follow where the assets and proceeds go.

New capabilities

• See NetLong from every angle. Research futures open interest and margin debt across dynamically discovered markets, inspect positions and liquidation conditions, and build a product summary spanning usage, fees, house economics and LOSE distributions. Keep exposure, revenue and LP/lender results distinct.
• Compare standard-bond epochs. Ask for the last ten completed epochs: NET sold, USDG received, opening NAV and market price, time until less than one NET remains, and period totals. Receipt reconciliation and explicit gaps keep partial history useful without presenting it as complete.
• Follow NET sales into Treasury. Measure PremiumSeller volume, realized weighted sale prices and receipt-confirmed USDG delivery, with clear separation between issuance proceeds and earned revenue.
• Investigate revenue by product and asset. Trace fees, interest, underwriting results and recipient flows across NetNet. Separate earned economics from Treasury/Sleeve receipts, financing and profit, without counting internal transfers twice.
• Understand backing and leverage. Break down Core RFV, the publisher's True RFV and adjusted net assets; investigate Sleeve holdings, off-wallet claims and borrowing, with fresh, block-specific reconciliation.
• Go deeper on product mechanics. Expanded InverseBond, Sleeve bonds, Credit/Loopback, Predict/House and THE BOOK coverage connects functions, accounting consequences and evidence. Research current announcements, interviews, documentation and UI claims alongside contracts.

Ask naturally. Current questions use fresh authorized evidence; missing history and unsupported conclusions stay explicit. The skill remains a research aid, not a wallet operator.

Verification and security

• 495 tests passed locally; all four final-commit CI jobs passed: Ubuntu 24.04/macOS 14 × Python 3.10/3.14. CI results.
• Exact-export CLI smoke checks exercised confirmed accounting, arithmetic/header mismatches and terminal denials. Runtime archives rebuilt byte-identically. These are synthetic checks, not fresh live-chain acceptance.
• Independently reviewed runtime code and security-reference bytes are unchanged from the reviewed preflight; final scan findings received current-context dispositions. AI-assisted reviews are not human audits or deployed-contract certification.

Automated scanner results are not all clean. Raw findings and individual dispositions are preserved in the separate audit archive.

Gate	Recorded result
Cisco Skill Scanner 2.1.0	Automated FAIL: 2 CRITICAL, 4 HIGH, 28 MEDIUM, 1 LOW. Six local analyzers; 313 files loaded and 21 complete Python contexts.
Hermes standalone SkillsGuard	Caution; default community-source installation blocked. 14 findings retained, including actual file-count and size-limit exceedances. No forced installation.
OpenClaw official standalone validator/packager	Passed; all 313 archive members match the frozen input.
Gitleaks 8.30.1 runtime/history	Raw exit 1 retained: 32 runtime and 40 history findings across 74 commits. All individually supported as public addresses, recomputed hashes or static provenance labels; no unresolved credentials.
Gitleaks public audit payload	105 frozen evidence files scanned; raw exit 1 retained. All 127 findings individually supported as 75 file fingerprints and 52 public-address occurrences; no unresolved credentials.

Cisco's intended networking, declaration-schema mismatch, file-count and reference-depth limitations remain disclosed. Critical/high lexical matches were reviewed in their actual accounting and safety context; no demonstrated vulnerability requiring a runtime change was identified. Native diagnostic cache warnings remain recorded. Complete file loading does not establish semantic traversal beyond the scanner's depth limit.

The public audit scan's own reports, final notes and CI records were appended afterward; no recursive scan of those additions or the final ZIP is claimed. Standalone scanner/packaging gates do not certify a consuming agent host. Review the exact revision and installation caveats before installing.

Assets and integrity

• netstack-0.5.0.zip — installable 313-file runtime package.
• netstack-audit-0.5.0.zip — review/scan evidence and limitations; not an installable skill.
• SHA256SUMS — checksums for both ZIPs. Download all three into one directory and run shasum -a 256 -c SHA256SUMS.

Runtime ZIP SHA-256:
ec19c15b3ffb1f1fbce9fcd49acd23abb2cb21743893920c3dfc38a9f1821abd

Manifest SHA-256:
1916afacb3ec4ad251429611e3e190d85e636ab20818295ed7bdf36edddd3dd9

Declared runtime digest:
753909636143bffca21dcdfa00ed36c580d1a85b3370dbf8731d31cc1be9b880

The annotated tag and checksums provide unsigned integrity evidence, not a publisher signature or authentication. Preserve the sibling export receipt outside the runtime tree when using the export workflow.