test(scope): differential regression suite proves agent-scope-hardening holds (Spec 105 PR H1)

[Dumbris]

Summary

Spec 105 PR H1 — scope-regression-suite part 2 (FR-011/013/014, SC-007, gaps FR01x-G4..G7). Closes tasks T107-T116. This is the FINAL phase of the nine-PR agent-scope-hardening epic (spec 105), started with PR A (#1279) and completed through B, D, E, H0, C (#1325), F (#1326) and G (#1328) — all already merged to main.

Its job is to prove, with a genuine two-fixture differential oracle, that every acceptance scenario in specs/105-agent-scope-hardening/spec.md (User Stories 1-3) actually holds against the already-shipped code, and that none of the pre-105 insecure test assertions the prior PRs inverted has crept back to its original form. This PR is test-only except for one small, genuine bug fix it found along the way (see below) — it does not change any production authorization logic.

Changes

• Two-fixture differential oracle (internal/server/scope_differential_test.go, new): newScopeFixture/runScopeScenario/normalizeScopeResponse per contracts/differential-oracle.md, plus TestScopeCoverage_EveryUserStoryScenario registering every US1.1-1.8/US2.1-2.6/US3.1-3.4 scenario by id — some directly via the general three-server harness, most by re-running each prior PR's own dedicated differential fixture (plan.md: "H1 re-registers them by User Story id").
• Pinned-reversal grep guard (internal/server/scope_pinned_reversal_guard_test.go, new): TestScopePinnedReversalsStayInverted proves none of the four pre-105 insecure assertions gap-map.md's FR01x-G7 entry names have crept back to their original form, traced to their exact pre/post-105 diffs across commits bac8f6b0d, 46de8038c, e33d6a139.
• Retained-effect fixtures (internal/server/scope_retained_effects_test.go, new): real reproductions of the SC-001-excluded, deliberately-unchanged effects — shared-limiter contention, cross-server scan admission (via the actual shadowing.cross_server impersonation-clone mechanism), prompt-name collision, global prompt cap, direct display-name collision, plus shared prompt-refresh deadline via a companion package (internal/upstream/manager_prompts_deadline_test.go, new, two discriminating regimes), and shared log rotation (cited cross-package).
• HTTP credential matrix (internal/server/scope_http_matrix_test.go, new): real minted tokens over real loopback HTTP through mcpAuthMiddleware, covering every FR-014 applicability-matrix row at least once plus a content-level (raw wire-text) non-disclosure proof for retrieve_tools.
• Latency harness finalized (internal/server/scope_latency_test.go): all four FR-011 operations (retrieve_tools, read_cache, prompts/list, tools/list) with real workloads, plus .github/workflows/scope-latency.yml (new) and cmd/scope-latency-compare (new, stdlib-only, unit-tested) for the merge-base p95 regression gate — verified end-to-end locally against this PR's actual merge-base commit.
• Bug fix found via this suite: internal/logs.ReadUpstreamServerLogTail/ReadUpstreamServerLogTailAttributed dereferenced a nil *config.LogConfig (affecting admin and scoped readers identically — a robustness gap, not a scope/security issue). Fixed with a TDD red/green pair (internal/logs/nil_log_config_test.go).
• Docs and roadmap sync: docs/features/agent-tokens.md invariant wording finalized (no overclaiming — most FR coverage is via re-run dedicated fixtures, not literally every FR through the general 3-server harness). roadmap.yaml's scope-refusal-shapes task was found still status: todo despite PR G (fix(scope): scope-first refusal precedence, hidden equals nonexistent on every surface (Spec 105 PR G, FR-010) #1328) having already merged — flipped to done with its PR number. scope-regression-suite's own task status and the epic-level agent-scope-hardening status are deliberately left as todo/in_progress for the maintainer to flip on merge, per this repo's "no self-merge" convention for these security PRs.

Test plan

• go build ./..., go vet ./... (bare + -tags server): clean
• Full -race suite across internal/server, internal/upstream, internal/cache, internal/logs, cmd/scope-latency-compare: green
• golangci-lint run --config .github/.golangci.yml (bare + --build-tags server, --new-from-rev=origin/main): 0 new issues
• gofmt -l clean on every touched/new file
• ./scripts/test-api-e2e.sh: 70/70 passed
• Cross-model review: opencode terra/sol/astra confirmed quota-exhausted before each round; fell back to codex exec --model gpt-5.6-sol --sandbox read-only per CLAUDE.md's fallback rule. 4 rounds, every finding verified genuine before fixing (never accepted a finding without first confirming it against the actual code, and every fix was verified non-vacuous by temporarily reverting the underlying guard/logic and confirming the new test goes red, then restoring it). Round 1 found 7 real issues across 8 review areas (including two genuinely vacuous retained-effect fixtures that never exercised the mechanism they claimed to); round 2 found 2 more (including the nil-pointer panic above) plus a fix that itself needed strengthening; round 3 found the deadline fixture still didn't rule out an independent-concurrent alternative design; round 4: VERDICT: clean — "Both prior MUST-FIX findings are resolved: The complementary 150ms/100ms fixture distinguishes shared-sequential behavior from both independent-concurrent and independent-sequential deadlines. The real-HTTP retrieve_tools test provides content-level non-disclosure through authentication middleware, with an authorized positive control. Combined with the differential suite and existing direct/profile HTTP checks, further per-row content assertions have insufficient marginal value." Full round-by-round history in specs/105-agent-scope-hardening/tasks.md T116.

Related #1223 #1224 #1225 #1226 #1227 #1279 #1282 #1283 #1284 #1285 #1325 #1326 #1328

[cloudflare-workers-and-pages]

Deploying mcpproxy-docs with    Cloudflare Pages

Latest commit:	446b25c
Status:	✅  Deploy successful!
Preview URL:	https://1d52b627.mcpproxy-docs.pages.dev
Branch Preview URL:	https://105-h1-regression-suite.mcpproxy-docs.pages.dev

View logs

[github-actions]

📦 Build Artifacts

Workflow Run: View Run
Branch: 105-h1-regression-suite

Available Artifacts

• archive-darwin-amd64 (30 MB)
• archive-darwin-arm64 (27 MB)
• archive-linux-amd64 (18 MB)
• archive-linux-arm64 (16 MB)
• archive-windows-amd64 (30 MB)
• archive-windows-arm64 (26 MB)
• frontend-dist-pr (0 MB)
• installer-dmg-darwin-amd64 (24 MB)
• installer-dmg-darwin-arm64 (22 MB)
• smart-mcp-proxymcpproxy-goOQ3K1P.dockerbuild (0 MB)

How to Download

Option 1: GitHub Web UI (easiest)

1. Go to the workflow run page linked above
2. Scroll to the bottom "Artifacts" section
3. Click on the artifact you want to download

Option 2: GitHub CLI

gh run download 35534699095 --repo smart-mcp-proxy/mcpproxy-go

Note: Artifacts expire in 14 days.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 61.53846% with 30 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
cmd/scope-latency-compare/main.go	57.14%	26 Missing and 4 partials ⚠️

📢 Thoughts on this report? Let us know!