Summary

Adds a contract test that keeps the platform support declarations named by the
roadmap Definition of Done in agreement: session-analysis CLI help and platform
gate, the exported SESSION_ANALYSIS_HELP, the agent-customize provider
registry, createAnalyzer rejection, harness analyze help and platform gate,
the asset-baseline provider gate, and the host adapter matrix. Any host added
or dropped in one owner without the others now fails npm test.

To let the test assert the report platform gate's whitelist (not just that it
rejects unknown input), the reportPlatform gate's rejection error now names
the full supported set, matching the session-analysis and asset-baseline gates.

Why

• Issue/Story: roadmap.md TODO A-06 (P0) — "Add support-declaration
  consistency tests."
• User or maintainer outcome: the supported platform set (qoder, codex,
  claude, cursor, qwen, copilot) is declared independently in at least
  five places. Nothing asserted they agree, so adding a host in one owner
  silently left CLI help, error messages, and docs advertising different
  platform sets (this actually happened when qwen and later copilot
  landed). This test closes that gap.

Traceability and Scope

• Spec/ADR, if applicable: docs/specs/2026-07-28-a06-support-declaration-consistency.md
• Acceptance criteria addressed: AC-1 (every surface's declared list equals the
  canonical set), AC-2 (public routes only: root facade CLI, the
  session-analysis/index.mjs and agent-customize/providers/index.mjs import
  surfaces, shipped docs), AC-3 (unsupported input fails closed naming the full
  set on each gated route), AC-4 (npm test discovers the test automatically).
• Canonical owners changed: scripts/harness-analysis/report-run.mjs —
  reportPlatform whitelist extracted to a file-local REPORT_PLATFORMS
  constant; its rejection error now names the supported set. Each capability
  still owns its own declaration. Also adds test/support-declarations.test.mjs,
  the spec, a CHANGELOG.md entry, and checks the A-06 roadmap box.
• Explicit non-goals: no change to which platforms are supported; no shared
  runtime constant across modules; ANALYZE_HELP keeps its own prose list
  (asserted separately by the test); no prose-wording assertions beyond the
  declared platform lists; other commands with their own platform/provider
  lists (coding-agent-practices inventory / asset-integrity,
  agent-customize, agent-lint, evidence-bundle, task-loop-source,
  selection-profile) stay out of scope.

Change Type

• Feature
• Bug fix
• Tests only
• Refactor with no intended behavior change
• Documentation/community
• Dependency, packaging, or infrastructure

Test and Review Evidence

Manual or visual evidence: n/a (no visual surface).

Risk and Recovery

• Compatibility and cross-platform impact: the only runtime change appends
  Supported platforms: ... to the reportPlatform rejection message; the
  unsupported Harness report platform: <name> prefix and the
  UNSUPPORTED_REPORT_PLATFORM error code are unchanged, and no other code or
  test depends on the old message format. Test uses process.execPath +
  spawnSync (no shell), so it runs on Windows, macOS, and Linux.
• Package, plugin, schema, or generated-file impact: none.
• Rollback or recovery path: revert the single commit; nothing depends on the
  new test file, the spec, or the extracted constant.
• Residual risk or unverified boundary: the test extracts lists from current
  help/error phrasing (--platform <...>, Supported platforms: ... .);
  rewording those messages requires updating the extraction patterns. Platform
  declarations in the out-of-scope commands listed under non-goals can still
  drift and are intentionally not covered.

AI Involvement

• Level: Generated (agent-authored; co-author recorded in the commit trailer).
• Human review and validation: maintainer reviewed the analysis, diff, and spec
  wording, approved each step; a code-review pass caught and fixed a spec
  statement that overstated help/constant sharing; the branch was rebased onto
  the latest main and the reportPlatform conflict resolved manually; the
  listed test commands were executed locally with results reported verbatim
  above.

Checklist

• I followed AGENTS.md, CONTRIBUTING.md, and the relevant canonical-owner guidance.
• The change is focused and does not include unrelated local or generated state.
• Tests and documentation match the behavior actually delivered.
• Markdown links were checked when documentation moved or changed.
• Cross-platform behavior was considered for Windows, macOS, and Linux.
• Package/runtime verification was run when shipped files or dependencies changed.
• User-facing or compatibility changes are recorded in CHANGELOG.md.
• I have the right to contribute this work under the repository's MIT License.