Two lines still cited the pre-A2 state (294 tests / 40 files,
"known gaps remain at save_scene / POST boundary"). Corrected
against the actual 302/41 and the Phase 10 A2 closure of those
boundaries via the shared apiGraphSchema.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
A5 wrote the initial PR description before the A2 security fixes
landed, so it cited 294 tests / 9 commits and listed several now-fixed
items as follow-ups. Updated:
- 294 -> 302 tests (8 new SSRF guard tests in safe-fetch.test.ts)
- 40 -> 41 test files
- 9 -> 20 commits with the actual Phase 5-10 series listed
- Security-notes section: document safeFetch + shared apiGraphSchema
wiring on both POST and PUT; remove "tracked follow-up" for
save_scene URL validation (done); add "tracked follow-up" keeps
same-id write race + auth layer
- Checklist: flip two items from [ ] to [x] for URL validation and
SSRF protection
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Five parallel audit agents reviewed the branch before open-sourcing
the PR to pascalorg/editor:
- a1-secrets.md: SAFE TO PUSH. Scanned 176 files / 40,768 diff lines.
Zero secrets, tokens, API keys, PEM blocks, JWTs, or cookies.
Only MEDIUM finding: absolute /Users/adrian paths in test-report
scripts (cosmetic, not security).
- a2-security.md: FOUND 2 HIGH-severity issues, both FIXED in
commit 8757de0:
* PUT /api/scenes/[id] still had the loose graphSchema that POST
got fixed in Phase 8 P4. Shared schema extracted to
apps/editor/lib/graph-schema.ts so both routes re-validate.
* photo_to_scene + analyze_floorplan_image + analyze_room_photo
all did raw fetch(url) on user-supplied URLs - a textbook SSRF
to 169.254.169.254 cloud metadata. Added safe-fetch.ts with
private-IP / link-local / .local-hostname denylists, manual
redirect revalidation, size cap, timeout, env-allowlist.
- a3-code-quality.md: READY FOR REVIEW. Zero production `any`, all
tools Zod-validated in+out, uniform error handling,
conventional-commits. Two non-blocking follow-ups: client editor
components (SceneLoader, SaveButton) have no tests; document
check_collisions n^2 scaling.
- a4-performance.md: SHIP WITH NOTES. MCP dist 904 KB, Supabase
lazy-imported (zero editor bundle impact), v0.1 hot paths
sub-200ms. Flagged: FilesystemSceneStore.index.json O(n) per
write (fine <1k scenes), concurrency races (documented in P8),
client render at 5k nodes unverified.
- a5-pr-description.md: polished final PR description that
corrected stale test counts (294 not 142), disclosed all 5
cross-cutting surfaces, named the known failures honestly,
split the checklist, expanded the scope to the real Phase 7
deliverables.
Overall verdict: READY TO PUSH after the A2 fixes landed. No
blockers remain.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Pre-push audit agent A2 flagged two HIGH-severity security issues that
would have shipped in the PR had we not checked:
1. PUT /api/scenes/[id] URL-validation bypass. Phase 8 P4 fixed
the POST /api/scenes route by replacing a loose z.unknown() graph
schema with AnyNode superRefine. The fix never made it to the PUT
handler - an attacker could resubmit the same javascript:/file:///
payloads via PUT. Fixed by extracting the tight validator into
apps/editor/lib/graph-schema.ts and sharing it across both routes.
2. SSRF in photo_to_scene / analyze_floorplan_image /
analyze_room_photo. All three tools called raw fetch(image) on
user-supplied URLs with no validation - a direct
http://169.254.169.254/latest/meta-data/ exfil primitive on any
cloud host. Added packages/mcp/src/lib/safe-fetch.ts that:
- Blocks loopback (127.0.0.0/8, ::1)
- Blocks link-local incl. cloud metadata (169.254.0.0/16)
- Blocks private ranges (10/8, 172.16/12, 192.168/16, fc00::/7)
- Blocks .local/.internal/.corp hostnames + localhost variants
- Blocks v4-mapped IPv6 loopback (::ffff:127.0.0.1)
- Manual redirects (max 3), revalidating the allowlist per hop
- 20 MB response-size cap (streamed, enforced per-chunk)
- 10s timeout
- Optional PASCAL_ALLOWED_ASSET_ORIGINS env allowlist
Tests: 8 new SSRF guard tests, all vision tests still pass, full
suite 302/302.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Seeds the shared scene store with Casa del Sol + 3 templates + 3 Villa
Azul variants so /scenes can be browsed end-to-end after a fresh
`bun dev` + `pascal-mcp --http` restart.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Builds a larger, richer house than Casa del Sol via MCP save_scene
(no injection hack), then dispatches 10 parallel verifiers across
schema, geometry, dimensions, openings, HTTP, page render,
parentage, round-trip, spatial, visual.
Villa Azul — 56 nodes, validate_scene=true, 44KB on disk:
- 15x10m building envelope (vs Casa del Sol's 12x8)
- 9 interior zones (master bed/bath, bed 2/3, shared bath,
living/dining, kitchen, entry hall, corridor)
- 10 doors + 12 windows (all cut successfully)
- 4 exterior zones (pool 8x4 + basin slab at -2m, outdoor kitchen,
driveway, back patio)
- 5 rail-style fences (vs Casa del Sol's privacy) with 2m entrance gap
Verification: 108 checks, 104 PASS, 4 findings:
- V1 schema: 56/56
- V2 geometry: 7/7 (perimeter closes, interior T-junctions, no
zone overlaps, fence gap verified)
- V3 dimensions: 13/13 zone areas exact (1 spec mismatch on site
polygon default, not a build bug)
- V4 openings: 22/22 dimensional fit, surfaced a tool gap in
cut_opening (no adjacency check) + my build packed too tightly
- V5 HTTP: 10/10 (GET/PUT/PATCH/DELETE/HEAD, If-Match conflicts)
- V6 page: 14/14 (/scene/:id 81KB, /scenes 20KB, 404 fallback)
- V7 parentage: surfaced CROSS_CUTTING §2 site->building->level
parentId=null (pre-existing in core's loadScene)
- V8 round-trip: 10/10 byte-equal, duplicate_level -> 110 nodes
- V9 spatial: 12/12 (find_nodes, measure, constraints resource)
- V10 visual: HTML fallback (Chrome extension disconnected during
run); API layer intact
Follow-up tracked: `cut_opening` should check opening-adjacency on
the same wall (minimum gap) to catch tight packing during patch
construction. Currently returns success and relies on the UI to
visualise the overlap.
Live at http://localhost:3002/scene/a6e7919eacbe.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Phase 8 parallel validation flagged two boundaries where malicious URLs
(javascript:, file:, external http:, data:text/html, ...) could be
persisted despite the AssetUrl allowlist added in Phase 7 A7:
1. `save_scene({ includeCurrentScene: false, graph })` — the graph arg
was treated as opaque (`z.record(z.string(), z.unknown())`) and
written to the store without re-running AnyNode.safeParse.
2. `POST /api/scenes { graph }` in the editor API — same issue; the
Zod `graphSchema` accepted anything object-shaped.
Fixes:
- `save-scene.ts`: when `includeCurrentScene === false`, iterate every
node and run `AnyNode.safeParse`; collect issues and throw
`McpError(InvalidParams, 'graph_invalid', { errors })` on any
failure.
- `app/api/scenes/route.ts`: replace `graphSchema` with a structured
`z.object({ nodes, rootNodeIds, collections? })` + `superRefine`
that runs `AnyNode.safeParse` on every node. Invalid → 400 with
detailed issue paths.
Tests:
- Added `save_scene` regression test for the P4 attack
(item.asset.src = 'javascript:alert(1)') — expected error.
- Fixed the existing `includeCurrentScene=false` test to use a
schema-compliant site node id (the prior `id: 'root'` now fails
the AnyNode parse, which is the desired strict behaviour).
- Full suite: 294 pass / 0 fail.
Also adds Phase 8 test-reports/phase8/** (10 agents, ~15 scripts +
markdown reports) documenting the validation run, plus minor biome
cleanups to the Phase 5/7 test artefacts (removed stale
`// biome-ignore` suppression comments that now resolve to the
already-off `noConsole` rule).
Phase 8 result summary (10 parallel agents, stdio MCP transport with
isolated data dirs):
- P1 templates: 18/18 PASS
- P2 variants: 6/7 mutations + determinism + save + combined + error
- P3 locking: 12/12 PASS (MCP + editor HTTP If-Match)
- P4 URL hardening: fixed 2 bypasses (see above)
- P5 photo-to-scene: 6/6 PASS
- P6 Casa del Sol via save_scene: 13/13 PASS
- P7 editor HTTP API: 18/18 PASS
- P8 concurrency: 4/5 PASS, flagged 2 real filesystem-store races
(expectedVersion CAS gap + .index.json drift under parallel writes)
- P9 edge cases: 13/13 PASS (size cap, slug safety, bad inputs)
- P10 full sweep: 37/37 PASS (30 tools + 4 resources + 3 prompts)
Known follow-ups:
- FilesystemSceneStore needs a proper lockfile / atomic CAS to fix
the P8 concurrency bugs (low priority: single-writer MCP is the
typical case).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
sudhir