Uvilo OS History
Archived Sessions (2026-03-07 to 2026-03-13)
Infrastructure build, Starlight site setup, Vercel deployment, sidebar/fonts/customization,
plugin cleanup, custom markdown plugins (remark-rehype), agentic execution spec, Zod v4 fix,
uvilo-git-remote backup, project-based workflow rewrite (History/WIP naming), instructions.md
rewrite for Railway/LibreChat, Playwright MCP (sidecar Chromium, auth pattern, env var fixes),
Vercel MCP, /git-status artifact dashboard (hosted Astro page + iframe artifact),
/preview command with build polling, MCP server installations (Context7, GitHub, Neon, Linear),
MCP verifications, deferred tools configuration, SurveyMonkey MCP research.
Key milestones: all 17 artifact dashboard items done, 12+ MCP servers verified, recursion limit increased to 100, project-folder naming convention established.
Archived Sessions (2026-03-13 to 2026-03-14)
uvilo-trash MCP implementation (6 tools, FastMCP), /git-status skill refactor (split into
Spec + Command docs), trash verification (5/6 tools passed, empty_trash(0) schema bug fixed
with Annotated[int, Field(ge=0)]), Railway MCP (npm scope hyphen fix), SuprSend MCP
(curl→urllib rewrite, tarfile filter fix, -w flag correction), PostHog verified.
Archived Sessions (2026-03-16 to 2026-03-18)
uvilo-shell design and implementation (single run() tool, quote-aware chain parser, two-layer
architecture, permissions system with once/session/always grants), shell verification (4/4 tests
passed), SuprSend schema proxy (draft-07→draft-2020-12 rewrite), SurveyMonkey MCP via Composio
Platform (migrated from deprecated mcp.composio.dev, connection fix, verified 5 tools), naming
normalization (folder name = file prefix rule), diverged branch resolution lesson.
Session 2026-03-18 — SurveyMonkey MCP Live + Baseline Survey Results
Summary: SurveyMonkey MCP was tested and confirmed fully working — 5 tools loaded (GET_GROUPS, GET_BULK_CONTACTS, LIST_BENCHMARK_BUNDLES, CREATE_SURVEY_FOLDER, DELETE_SURVEY), all returning HTTP 200. Listed 5 surveys in the account, then fetched and decoded the full Baseline Survey results (3 respondents, 26 questions), presenting them in an interactive React artifact with Responses, Summary, and Demographics tabs. Also added a “Composio MCP” todo to the MCP section of State.
State changes: SurveyMonkey MCP changed from [BLOCKED] → [x]. Composio MCP remains [ ] (separate task, not yet defined/completed).