docs: link published IETF draft, refresh IPP comparison to draft-01; fix(deps): clear 10 Dependabot advisories

[asiridalugoda]

Two things: documentation updates for the published IETF draft, and a lockfile-only dependency triage.

Part 1 — Documentation

Two things changed upstream since these docs were written:

• draft-helixar-hdp-agentic-delegation-01 is now published on the IETF datatracker (posted 2026-08-03, expires 2027-02-04).
• draft-haberkamp-ipp-01 superseded ipp-00 in July 2026.

README

• Add an IETF Internet-Draft badge and a Standardization note. The root README had no link to the draft at all, while every package README (hdp-langchain, hdp-crewai, hdp-autogen, hdp-agent-framework, hdp-autogen-ts) already links it.
• Clarify max_hops in the quickstart example and the verification pipeline: the value is the issuer's delegation budget, not a protocol limit, and omitting it leaves chain length unbounded. This mirrors wording published in -01, prompted by a reader who read the example value 3 as a protocol-fixed three-agent cap.
• Point the "Why Not IPP?" link at ipp-01.

Package READMEs are deliberately not touched: they link the version-less datatracker URL, which always resolves to the newest revision.

COMPARISON.md

Every IPP claim was re-verified against draft-haberkamp-ipp-01 rather than assumed to carry forward:

Claim	Status in -01
5,000 ms revocation polling	Unchanged (8.2)
Genesis seal bound to khsovereign founding key	Unchanged (5.2)
Mandatory W3C DID Core principals	Unchanged (6.2)
Central taxonomy registry	Unchanged (10.3)
Proof-of-humanity absent	Still absent
Token structure	Changed
Hop signing	Changed

So two cells were corrected, not just requoted:

1. Token structure — genesis_seal renamed to genesis, intent_envelope to intent, and token_id / delegation / schema_version added (4.1-4.2).
2. Hop signing — provenance records now carry a per-record agent_sig (9.2). "Hop-level signing semantics are not specified" was true of -00 and is no longer accurate. Narrowed to the claim that still holds: the canonical serialization used for signing, and whether records chain cryptographically over prior records, remain undefined.

The core architectural argument (fully offline verification, no central trust anchor) is unchanged.

Part 2 — Dependency triage

package-lock.json only. No package.json ranges changed, so no API or supported-version surface moves. 10 of 11 open advisories cleared.

Runtime (ships to consumers)

• fast-uri 3.1.0 → 3.1.5 — clears four high-severity advisories: path traversal via percent-encoded dot segments (GHSA-q3j6-qgpj-74h6), host confusion via percent-encoded (GHSA-v39h-62p7-jpjc) and literal-backslash (GHSA-v2hh-gcrm-f6hx) authority delimiters, and host confusion via failed IDN canonicalization (GHSA-4c8g-83qw-93j6). Reached through ajv → ajv-formats, which src/schema/validator.ts registers — so these parse untrusted token input during verification. This is the group worth caring about.
• uuid 13.0.0 → 13.0.2 — GHSA-w5hq-g745-h8pq, missing buffer bounds check in v3/v5/v6 when buf is provided. Not reachable here: the only call site is uuidv4 in src/token/builder.ts, with no buf argument.

Development only

• postcss → 8.5.25 (GHSA-6g55-p6wh-862q, GHSA-r28c-9q8g-f849, GHSA-qx2v-qp2m-jg93)
• vite → 8.2.0 (GHSA-fx2h-pf6j-xcff, GHSA-v6wh-96g9-6wx3)

Left open, deliberately

• esbuild 0.27.4 (GHSA-g7r4-m6w7-qqqr, low). Patched in 0.28.1, but tsup@8.5.1 — the latest release — requires esbuild ^0.27.0, so reaching the patch would mean overriding tsup onto an unsupported esbuild major. The advisory covers arbitrary file read from the esbuild dev server on Windows; esbuild is used here only as a bundler via tsup and via vite for vitest, and no dev server is run. Recommend leaving it until tsup widens its range.

Verification

• 84 tests pass across 24 files (npm test)
• tsc --noEmit clean
• tsup build succeeds for CJS, ESM, and DTS

Known gaps

• IPP verification was done against fetched renderings of the -01 text, not a full line-by-line read. The rows above are solid, but a complete pass may surface more drift; the new delegation object in particular is unexamined. Worth closing before COMPARISON.md is cited in any announcement.
• https://helixar.ai/labs/hdp currently returns 404. Resolved — it now 301s to /about/labs/hdp/ and returns 200, so the link in COMPARISON.md works. Note for -02: that destination is an overview page, while the published -01 cites it as [HDP-SPEC] titled "v0.1 Specification" (§1.4: "The full HDP protocol specification is available at..."). Either publish the spec at that URL or retarget the reference.
• The vite 8.2 bump surfaces a new non-fatal warning: vitest.config.ts uses ESM syntax in a file loaded as CommonJS, which a future Vite major will stop supporting. Not addressed here; fix is to rename to .mts or set "type": "module".