chore: release package

[github-actions]

This PR was opened by the Changesets release GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.

Releases

@adcp/sdk@7.1.0

Minor Changes

• 7de0872: fix(comply): autodetect request_signer and oauth_metadata prereqs (comply(): autodetect request_signer and oauth_metadata requirements (mirror webhook_receiver autodetect from #1678) #1702)

  Extends the implicit-requires pattern shipped in comply() sends relative push_notification_config.url — 3.0-strict sellers reject #1678 (webhook_receiver)
  to two more requirement surfaces, removing ~44 false-negative step
  failures per comply() run on agents that don't claim the underlying
  capability.

  • Adds 'request_signer' to RequirementName / KNOWN_REQUIREMENTS /
    REQUIREMENT_TO_SKIP_REASON and extends detectImplicitRequires to
    flag any storyboard whose id === 'signed_requests' or that contains
    a request_signing_probe step. The gate consults the agent's
    get_adcp_capabilities.request_signing.supported from the discovered
    profile (not a runner opt-out flag) so an agent that claims the
    capability but hasn't pre-registered the runner's compliance test
    keypair still fails — matches the universal storyboard's stated
    gating contract at
    compliance/{version}/universal/signed-requests.yaml ("absence of
    advertisement is not a failure").
  • Pre-empts oauth_discovery cascade-skip on security_baseline when
    the agent's capabilities don't declare account.authorization_endpoint.
    The existing reactive phaseAbsent cascade (Runner should unconditionally probe RFC 9728 PRM regardless of test-kit declaration #677) only triggers on a
    404 from /.well-known/oauth-protected-resource; non-404 responses
    from a bearer-only agent (200-HTML, 405, redirect — the Wonderstruck
    shape) fell through to validation failures inside an optional: true
    phase. The pre-empt is phase-level, not storyboard-level, so the
    universal unauth_rejection and mechanism_required checks still run.

  Part of the coordinated stance at Tracking: ship 1676–1680 as one stance — the SDK is a witness, not a translator #1685 (the SDK is a witness, not a
  translator).

• e1ec3ef: Add structured notices: RunnerNotice[] advisory surface to StoryboardResult and ComplianceResult (adcp-client#1704).

  • New types: RunnerNotice, NoticeCode, NoticeSeverity — exported from @adcp/sdk
  • StoryboardResult.notices is always-present (default []); ComplianceResult.notices is optional, deduplicated by code across all storyboard runs
  • Two spec-grounded notices emitted on day one:
    • request_signing_required_in_4_0 (future_required) — when the signed_requests storyboard runs against an agent that lacks request_signing.supported: true (signing required for spend-committing operations in AdCP 4.0)
    • legacy_hmac_fallback_removed_in_4_0 (deprecation) — when agent declares webhook_signing.legacy_hmac_fallback: true (removed in AdCP 4.0)
  • CI gates and JUnit consumers can now key on stable code values for badge routing instead of parsing prose skip.detail strings

• 36d279f: fix(runner): attribute Zod schema rejects to response_schema, short-circuit invariants (Harness error attribution: Zod validation rejects surface as failures on unrelated downstream assertions #1709)

  When the SDK's response unwrapper rejects an agent response against the
  codegen-emitted Zod schema for the tool (e.g. an additionalProperties: true field the runner's .strict() schema doesn't enumerate, like the
  recently-added authorization field on sync_accounts), the exception
  previously propagated as a generic Error with a freeform message.
  The runner's step-execution catch absorbed it into stepResult.error
  but never attributed the failure to schema validation in
  step.validations[]. Step-scope invariants then ran against the
  malformed payload, and whichever fired first (canonically
  context.no_secret_echo — the next assertion in the default-invariant
  queue) became the surfaced failure. BidMachine spent 10+ deploys
  chasing the no_secret_echo ghost before the strict-Zod root cause
  surfaced. Full trace: comply runner: storyboard verdicts contradict spec-valid prod probes across 10+ runs adcp#4419.

  Fix:

  • New ResponseSchemaValidationError typed error class in
    src/lib/utils/response-unwrapper.ts. Carries toolName, issues
    (raw Zod issues), and data (the rejected payload) for downstream
    attribution. Stable name === 'ResponseSchemaValidationError' so
    cross-bundle consumers can detect by string match without instanceof.
    Replaces the two generic new Error('Response validation failed for ${toolName}: ...') throws in the unwrapper.

  • runStep in src/lib/testing/client.ts now returns
    { result?, step, caughtError? }. The new caughtError is the
    raw thrown value (typed unknown) so callers can pattern-match on
    typed exceptions. Backwards-compatible — pre-existing callers that
    consume only result and step are unaffected.

  • Storyboard runner executeStep threads caughtError from the
    dispatch fn, and on ResponseSchemaValidationError prepends a
    synthesized response_schema ValidationResult to step.validations.
    The synthesized entry carries the structured issues, the failing
    tool name, an RFC 6901-shaped json_pointer, and the rejection
    message. extractFailures.find(v => !v.passed) now picks it up
    before any inline or invariant entry.

  • Step-scope invariants in executeStoryboardPass are short-circuited
    when the synthesized response_schema entry is present. Each bypassed
    invariant emits a single skip marker entry (passed: true) so consumers
    see WHICH invariants were bypassed and why, but the bypass entries
    don't crowd out the schema failure in extractFailures.

    Sequencing: this PR is the upstream dependency for Schema URLs and Zod validators are baked at SDK build time — drift from agent's advertised adcp_version #1707 (dynamic
    schema fetch + strict→passthrough flip + codegen regen). Lands first
    so any remaining Zod rejects during Schema URLs and Zod validators are baked at SDK build time — drift from agent's advertised adcp_version #1707's rollout produce honest
    signal instead of misattributing to no_secret_echo. After both Harness error attribution: Zod validation rejects surface as failures on unrelated downstream assertions #1709
    and Schema URLs and Zod validators are baked at SDK build time — drift from agent's advertised adcp_version #1707 land, BidMachine's 63/128 comply-vs-45/59 CLI delta
    (adcp#4419) should collapse — fgranata volunteered as the retest
    target on adcp-client#1711.

    Coordinated stance: adcp-client#1685 (the SDK is a witness, not a
    translator). Same anti-pattern: the SDK was fabricating a different
    failure (no_secret_echo) than what actually went wrong (schema
    rejection); this PR makes it a faithful witness.

Patch Changes

• 05ce456: fix(version): auto-derive COMPATIBLE_ADCP_VERSIONS from ADCP_VERSION pin

  The 3.0.x patch enumeration in COMPATIBLE_ADCP_VERSIONS was a hardcoded
  array literal inside the scripts/sync-version.ts template. Every AdCP
  patch bump needed someone to remember to append the new version to it.
  The 3.0.9, 3.0.10, and 3.0.11 chore PRs all forgot — the list capped
  at 3.0.8 even though ADCP_VERSION moved to 3.0.11. Symptom:
  isCompatibleWith('3.0.11') === false against the SDK's own pin.

  Same root-cause class as the schema URL pinning drift surfaced by
  adcontextprotocol/adcp#4419 (BidMachine reports / "3.0.1 schemas" cited
  against a 3.0.11 seller): a load-bearing version surface that depends on
  human discipline at every patch bump.

  Fix:

  • scripts/sync-version.ts now derives the list dynamically from the
    current ADCP_VERSION. Enumerates 3.0.0..3.0.<patch> mechanically;
    the bumper no longer has to remember anything.
  • Fails closed when ADCP_VERSION falls outside the 3.0.x range so a
    future 3.1.x or 4.x bump forces the script to be extended (rather than
    silently inheriting a stale enumeration). The compat surface for a
    major/minor move is rarely mechanical — that's the right time to think.
  • Adds test/lib/compatible-versions-self-consistency.test.js asserting
    the regenerated list contains the current pin and fills the
    3.0.0..ADCP_VERSION range without gaps. Future regressions (someone
    reverting to hardcoded literals) fail loud at CI.

  Does not address the deeper schema URL pinning drift in
  src/lib/testing/storyboard/validations.ts (SDK-build-time ADCP_VERSION
  used in cited schema URLs regardless of the agent's advertised version);
  tracked separately at adcp-client#NNNN.

• 110b49e: fix(invariant): no_secret_echo only fails on string-valued suspect-named fields (no_secret_echo invariant flags spec-legitimate field names on structured-value fields #1713)

  The default context.no_secret_echo invariant flagged any response field
  whose name was in SUSPECT_PROPERTY_NAMES ('authorization', 'api_key',
  'apikey', 'bearer', 'x-api-key') regardless of the field's value.
  This over-rejected spec-legitimate structured fields — notably the
  authorization object on validation-result.json (a structured
  authorization-validation payload, not a credential) and any seller-side
  extension fields named authorization under
  sync_accounts_response.accounts[] (additionalProperties: true).

  Symptom: BidMachine's sync_accounts failures in
  comply runner: storyboard verdicts contradict spec-valid prod probes across 10+ runs adcp#4419 all surfaced as context.no_secret_echo.
  Diagnosis in #4419 pointed at Zod .strict() codegen lag, but
  verification of the published SDK tarballs (5.25.1 / 6.12.0 / current)
  shows SyncAccountsResponseSchema.accounts[] already uses
  .passthrough(). Zod silently accepts the authorization field; the
  NAME dragnet in this invariant is the actual rejection.

  Fix: narrow findSecretEcho so the suspect-name check only fires when
  the field VALUE is a non-empty string. Structured object/array values
  on suspect-named fields pass through to the recursive walk, which still
  scans nested strings against BEARER_TOKEN_PATTERN and caller-supplied
  secret literals. The actual leak shapes the invariant was designed to
  catch (bearer tokens, API keys, caller secret echoes) remain caught.

  Adds five new test cases covering:

  • Suspect name with object value passes (the BidMachine case)
  • Suspect name with array value passes
  • Suspect name with empty string passes (no leak)
  • Suspect name with non-empty string value still fails (existing dragnet)
  • Bearer literal nested inside a structured suspect-named object still
    fails via the value-scan regression guard

  Sequencing relative to other in-flight work:

  • adcp-client#1709 (PR fix(runner): attribute Zod schema rejects to response_schema (#1709) #1712, merged) — Zod-reject error attribution.
    Addresses a different misattribution path (when Zod DOES reject).
    This fix addresses the case where Zod ACCEPTS but the invariant
    over-rejects on field name.
  • adcp-client#1707 — dynamic schema fetch + strict→passthrough +
    codegen regen. Real architectural cleanup; does NOT unblock
    BidMachine (the published SDK already uses passthrough).
  • adcp-client#1711 (fgranata) — BidMachine's report. The "Zod
    .strict() codegen lag" diagnosis was incorrect for the published
    SDK; this fix is the actual unblocker. Closes once fgranata retests.

  Coordinated stance: adcp-client#1685 (the SDK is a witness, not a
  translator). Same anti-pattern: the SDK fabricated a credential-leak
  signal against a structured non-credential field that the spec legally
  permits.