Repository navigation
Conversation
Implement the published navigation and companion intent-delivery drafts across types, domain bindings, SDK exports, runtime injection, manifest inference, and the reference conformance runtime. Host-verified manifests determine retained invocation eligibility; traversal correlates reports with target-scoped delivery IDs. Constraint: Navigation depends on the companion published INTENT delivery draft; its closed PR remains intended by the maintainer. Rejected: Napplet-owned history snapshots or browser History API interception | Neither belongs to the current draft contract Confidence: high Scope-risk: moderate Directive: Preserve endpoint lifetime, report high-water ordering, and delivery-to-visit association when updating these drafts Tested: Full build (12 tasks), type-check (19 tasks), recursive unit suites, real Chromium manifest-bound navigation, JSR exports, skills contracts Not-tested: Production shell implementations outside this repository Co-Authored-By: Codex <codex@openai.com>
Document committed-location reporting, runtime-owned history, optional-domain availability, report recovery, and target-scoped delivery correlation. Record verification and review evidence alongside the quick-task plan. Confidence: high Scope-risk: narrow Tested: Documentation build, 26 assembled-site links, 11 skills contracts, strict TypeScript check of reporting example Directive: Recheck the linked living drafts before changing interoperability guidance Co-Authored-By: Codex <codex@openai.com>
Record completion against the open PR and preserve the checked source revision for future protocol updates. Confidence: high Scope-risk: narrow Tested: Open PR #280 and passing full build, types, 881 units, export and documentation checks recorded in task summary Co-Authored-By: Codex <codex@openai.com>
Merge main at f28c2fd so navigation sits on the contracts main introduced since merge base f683914. Transport and identity come from main; domain lists and export maps keep both navigation and llm. Fits that a text resolution would miss: - catalog: add a navigation entry, classified exploratory, because NAP_IMPLEMENTATIONS is keyed by every NapDomain - docs: add navigation to NAP_DOC_GROUPS, which drives the sidebar and index, and move the PR's intent delivery wording onto apps/docs/naps/intent.md - boot: carry the verified endpoint over the hello port with shell.handleFrom(endpoint, envelope) instead of the default endpoint - handler: run the navigation branch first, then main's domain filter, so correlated LLM invalid-request answers stay intact - tests: move the navigation shim tests onto main's fake port shell, and give same-id LLM and maturity fixtures distinct ids because the endpoint registry keeps the first binding for an id The PR's IntentDelivery.id, nullable sender and delivery-id-keyed controls are carried through unchanged; they implement earlier draft revisions. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…points These gaps exist only because main's host policy and the PR's endpoint registry now meet. - hasDomain intersects endpoint grants with the host ceiling, so a host with a narrow ceiling no longer serves history or delivers intents that it withheld - revoking a port binding ends the endpoint before the host removes the frame, in both attachReferenceShell and the boot harness, so a replaced document can't inherit the prior document's history or deliveries - detach() stays host teardown and leaves the endpoint alive - changeset notes that navigation is exploratory and needs a named opt-in Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Both sides of the merge added an import from @napplet/conformance to server.ts. It type-checked but dropped the aislop score below its 98/100 baseline. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Merged The merge needed a few changes beyond the conflict hunks:
Mergeable doesn't mean ready to land, though. Both drafts moved on 2026-10-31, after this was built, and the current text contradicts part of what's here:
I left all of that as it was, since this was only meant to resolve conflicts. Mind if we reconcile it with the current drafts before merging? The changeset and docs also say this follows the published drafts, so they'd change too. I also haven't checked the new rule that rejects Nit: the description says #128 is closed, and Not proven: 🤖 Generated with Claude Code |
Napplets can now report committed advertised intent locations and read runtime-owned history through
navigation.report/get. The reference runtime assigns visits, enforces manifest parameter allowlists and revision ordering, and revisits the exact retained entry through a correlated intent delivery to the same live instance.Targets the published NAP-NAVIGATION draft and companion INTENT delivery draft. The maintainer confirmed the companion remains intended despite its closed/unmerged PR. Breaking pre-1.0 change:
IntentDelivery.idis required;sender: nullidentifies shell-originated delivery. Runtime integrations must implement the same draft contract.Validation: full build (12 tasks), type-check (19 tasks), recursive unit suites (881 passing; 3 existing ignored), JSR exports, 11 skills contracts, and assembled-site crawl (26 URLs) pass. Includes 224 conformance tests and real Chromium manifest-bound navigation coverage. Independent review findings for reset lifetime and sparse-array allocation were fixed with regressions. Pinned AI-slop scan is 98/100 with no code findings; only the unchanged transitive
deepmergelow advisory remains, with no fix available.No external production shell testing, merge, or publication included.