nova

This skill should be used when the user wants to visualize data. It intelligently selects the most suitable chart type from 26 available options, extracts parameters based on detailed specifications, and generates a chart image using a JavaScript script.

Conduct multi-round deep research on any GitHub Repo. Use when users request comprehensive analysis, timeline reconstruction, competitive analysis, or in-depth investigation of GitHub. Produces structured markdown reports with executive summaries, chronological timelines, metrics analysis, and Mermaid diagrams. Triggers on Github repository URL or open source projects.

End-to-end smoke test skill for Nova. Guides through: 1) Pulling latest code, 2) Docker OR Local installation and deployment (user preference, default to Local if Docker network issues), 3) Service availability verification, 4) Health check, 5) Final test report. Use when the user says "run smoke test", "smoke test deployment", "verify installation", "test service availability", "end-to-end test", or similar.

Master skill for all Nova/DeerFlow agent work. Use this skill whenever you are building full-stack apps WITH Nova's computer agent (scaffold → verify → green loop), working ON Nova itself (middleware, harness, sprint patterns, journal hooks), or diagnosing Nova's agent-computer pipeline (white preview, 409, crash-loop, panel bugs, Stop not working). Routes to specialized references: build-loop (P1–P5 deterministic build), maintain (harness/middleware/sprint), diagnose (pipeline regressions). Triggers on: 'build me a...', 'I'm getting a 409', 'white preview', 'browser tab is blank', 'stop not working', 'add middleware', 'new sprint', 'harness boundary', 'extend Nova', 'run dev_verify', 'why is the agent stuck', 'the panel is clipped', 'BUILD_JOURNAL', 'drive to green'. Use this skill proactively — if the task touches Nova's codebase or uses Nova to build something, this skill applies.

Run Nova locally end-to-end (Docker Compose + PM2 + dev-entrypoint.sh). Use when the user says "how do I run this", "start Nova", "make dev", "spin it up locally", or "I just cloned the repo". Goes through pre-flight (Docker, PM2, uv, pnpm checks) → `make dev` → nginx :2026 + gateway :8001 + frontend :3000 → verification (curl `/`, curl `/health`, curl `/workspace/chats/<id>`). The canonical local dev path: `make dev` starts `docker compose -f docker/docker-compose-dev.yaml -f docker/docker-compose.dood.yaml up` under PM2 supervision. Triggers on phrases like "run Nova locally", "make dev", "I just cloned the repo", "first time setup".

Author or wire a new hook emission in Nova's LangChain agent loop. Use when adding observability for a new tool, when promoting a logged audit event to a structured receipt, or when extending the agent's self-evaluation storage. Nova's hook surface today is `RunJournal` (`backend/packages/harness/deerflow/runtime/journal.py:38`) which implements `BaseCallbackHandler` (on_tool_start / on_tool_end / on_chain_start / on_chain_end); there is no separate `PreToolUse`/`PostToolUse` enum, so any new hook emission must go through `RunJournal.record_middleware` or a new `BaseCallbackHandler` subclass. Triggers on phrases like "add observability for X", "log when tool Y runs", "instrument Z", or before shipping any new middleware.

Diagnose and fix a loop-detection firing in Nova's agent run. Use when the user reports "the agent keeps doing the same thing", "agent is stuck in ENOENT hell", or "we hit loop detection again" — also when reading the run's `[self-test]` lines shows repeated identical tool calls or ENOENT-style grep/searches. Triages across Layer 1 (hash-based), Layer 2 (per-tool frequency), Layer 3 (dead-end search divergence) per `backend/packages/harness/deerflow/agents/middlewares/loop_detection_middleware.py:837` and recommends the next code change. Triggers on phrases like "loop fired", "agent is repeating", "ENOENT pattern", or after observing a loop in the Activity tab.

Edit the lead agent's system prompt with discipline. Use when the user says "the agent is being too X", "tell the agent to Y", "the prompt needs a rule about Z", or before adding any new rule to the system prompt. The system prompt is at `backend/packages/harness/deerflow/agents/lead_agent/prompt.py` (a single function that returns a `ChatPromptTemplate`). Use this skill BEFORE making the change; the skill walks through the change-and-validate cycle. Triggers on phrases like "edit the agent prompt", "add a rule to the lead agent", "system prompt", "AGENT_MANIFEST", or before merging any change touching `agents/lead_agent/prompt.py`.

Author a typed post-run receipt for a new Nova agent run or a new tool. Use when adding observability for an end-to-end workflow, when a tool produces artifacts a judge could re-verify, or when the "code is the ultimate proof" thesis needs a new structured artifact. The `Receipt` class doesn't exist yet (no hit on `class Receipt` in `backend/`); today the closest analog is `RunEventStore` (JSONL of `message/trace/lifecycle` events) plus `BUILD_JOURNAL.md` (free-form text) plus `REVIEW.md` (deterministic code-review). The schema is Pydantic v2 in `backend/packages/harness/deerflow/receipts/types.py` (planned v7.4). Triggers on phrases like "make this verifiable", "replay this run", "what did the agent do", or before shipping any new end-to-end workflow.

Verify a Nova change against the existing replay E2E contract. Use when changing any code that touches the agent loop, middlewares, tools, gateway routes, or sandbox behavior. Re-runs `backend/tests/test_replay_golden.py` (Layer 1, no API key) and the fullstack Playwright render (Layer 2, no API key). If replay misses, the fixture is stale and the skill walks through re-recording with `backend/scripts/record_gateway.py`. Triggers on phrases like "verify this", "is this safe", "does it break replay", or before merging any PR touching `backend/app/gateway/**`, `backend/packages/harness/**`, or `frontend/**`.

Debug a Nova sandbox failure (Local vs AIO) — which provider is active, why `browser_check` / `dev_verify` / `save_skill` are skipping, and how to flip providers. Use when the user reports "browser_check failed: requires the container sandbox", "save_skill wrote to wrong path", or "dev_server can't bind port 4100". Routes through `is_local_sandbox` (`sandbox/tools.py:1204-1221`), the LocalSandboxProvider vs AioSandboxProvider switch (`sandbox_provider.py:60-74`), and the AIO shell-session recovery path (`community/aio_sandbox/aio_sandbox.py:131-148`). Triggers on phrases like "sandbox won't start", "browser tool doesn't work", "save_skill went to host not container", or after observing `reason: "this sandbox has no browser"` in a `BrowserCheck` response.

Configure per-tool-type model routing in Nova to save 40-60% of LLM tokens. Use when configuring `config.yaml` to use a cheaper model for `browser_check`, `code_review`, `system_probe`, `save_skill` while keeping the lead agent on `opus`. The `ModelRouter` class doesn't exist yet (per the audit: no matches for `tool_type|model_routing|model_for_tool`); today every tool call uses the same model binding as the lead agent. This skill teaches the v7.4 pattern: a declarative `models.routing:` block in `config.yaml` + a `router.route(tool_type) -> ChatModel` resolver that runs once at `make_lead_agent` graph build time. Triggers on phrases like "cut token costs", "use haiku for screenshots", "route browser_check to a cheaper model", or after observing token-usage breakdowns via `SubagentTokenCollector` snapshots.

Use this skill when the user uploads Excel (.xlsx/.xls) or CSV files and wants to perform data analysis, generate statistics, create summaries, pivot tables, SQL queries, or any form of structured data exploration. Supports multi-sheet Excel workbooks, aggregation, filtering, joins, and exporting results to CSV/JSON/Markdown.

Use this skill when the user requests to generate, create, imagine, or visualize images including characters, scenes, products, or any visual content. Supports structured prompts and reference images for guided generation.

Use this skill when the user requests to generate, create, compose, or produce music or songs — background music, theme songs, jingles, or instrumental tracks. Generates a song from a style/mood prompt and optional lyrics via the MiniMax music API.

Use this skill when the user requests to generate, create, or produce podcasts from text content. Converts written content into a two-host conversational podcast audio format with natural dialogue.

Use this skill when the user requests to generate, create, or make presentations (PPT/PPTX). Creates visually rich slides by generating images for each slide and composing them into a PowerPoint file.

Use this skill when the user wants a systematic literature review, survey, or synthesis across multiple academic papers on a topic. Also covers annotated bibliographies and cross-paper comparisons. Searches arXiv and outputs reports in APA, IEEE, or BibTeX format. Not for single-paper tasks — use academic-paper-review for reviewing one paper.

Use this skill when the user requests to generate, create, or imagine videos. Supports structured prompts and reference image for guided generation.

Interact with DeerFlow AI agent platform via its HTTP API. Use this skill when the user wants to send messages or questions to DeerFlow for research/analysis, start a DeerFlow conversation thread, check DeerFlow status or health, list available models/skills/agents in DeerFlow, manage DeerFlow memory, upload files to DeerFlow threads, or delegate complex research tasks to DeerFlow. Also use when the user mentions deerflow, deer flow, or wants to run a deep research task that DeerFlow can handle.

Use this skill to review code changes for bugs, security issues, and quality. Trigger on "review this", "code review", "review my changes", "review the PR", or before shipping a feature. Reviews the working diff in the sandbox, comments by file with severity, and proposes concrete fixes.

Use this skill to organize a folder. Trigger on "organize this folder", "clean up these files", "sort my files", "propose a folder structure". Scans a directory in the sandbox, groups files by topic/type, and proposes a new structure — moving files only after approval.

Use this skill to test the user's current project. Trigger on "test this project", "run the tests", "qa", "check if it works", "find bugs", or after building a feature that needs verification. Runs the project's real test suite in the sandbox, captures failures, and proposes fixes as diffs (waiting for approval before writing).