ODS is a system for turning a PC, Mac, or Linux computer into a private local AI server, meaning a machine that runs models and related AI services itself. It supports local model inference, chat, voice, agents, workflows, document retrieval, search, and image generation. The catalogue includes commands and an instruction for operating or configuring this system.
Borrowing it
Nothing to install: this file belongs to Osmantic/ODS. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/Osmantic/ODS/main/.claude/commands/deep-research.mdgit clone --depth 1 https://github.com/Osmantic/ODSWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/osmantic/ods/deep-research)<a href="https://agentmods.dev/commands/osmantic/ods/deep-research"><img src="https://agentmods.dev/badge/commands/osmantic/ods/deep-research/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/osmantic/ods/deep-research"><img src="https://agentmods.dev/badge/commands/osmantic/ods/deep-research.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00021 | $0.05732 |
| Opus 5.5 | $0.00008 | $0.02293 |
| Sonnet 5.5 | $0.00004 | $0.01146 |
| Haiku 4.5 | $0.00002 | $0.00573 |
Grade A, and why
deep-research scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 620 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Deep Research with Multi-Model Consensus
Conduct comprehensive research on any topic by combining real-time web search (via Rube/Composio MCP) with multi-model deep analysis and consensus synthesis (via PAL MCP). Produces a structured research report with sourced findings, cross-validated analysis, and confidence assessments.
Arguments
$ARGUMENTS- Research configuration:- First positional argument: The research topic, question, or area of investigation
--depth <level>- Research depth:shallow(quick overview, 2-3 searches),medium(balanced, 5-7 searches, default),deep(exhaustive, 10+ searches with follow-up queries)--output <filepath>- Save the final report to a file (default: print to console)--models <N>- Number of models for consensus analysis (default: 3, min: 2, max: 5)
Philosophy: Ask Early, Ask Often
This skill should liberally use AskUserQuestion at every decision point. Research is inherently exploratory — assumptions about what the user wants are frequently wrong. The cost of asking is low; the cost of researching the wrong angle is high. Specifically:
- Before searching — confirm the research plan and sub-questions
- When the topic is ambiguous — clarify intent, scope, and angle
- After initial searches — share what was found and ask about direction
- When gaps are identified — let the user prioritize which gaps matter
- When contradictions surface — present both sides and ask for guidance
- When models disagree — let the user break the tie
- Before finalizing — confirm the report meets the user's needs
- After delivery — ask about follow-up research
The user should feel like a research partner steering the investigation, not a passive recipient of a pre-baked report.
Workflow
Phase 1: Parse Arguments and Plan Research
Extract the research topic from $ARGUMENTS. Parse optional flags:
- Default:
depth=medium,models=3, no file output - Identify the core question and decompose it into 3-7 sub-questions that, when answered together, provide comprehensive coverage.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2mo ago First seen · 620 lines · 21 tokens per session scan A f4c2a4fc1bc3
deep-research is a command published in the GitHub repository Osmantic/ODS (7,712 stars, last pushed today), licensed Apache-2.0. It adds 21 tokens to every session and 5,732 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other commands, from other repositories
review
Review code changes against project conventions.
resume-workflow
Restore the original Moira execution and continue its current step.
save-to-md
Save session documentation with Neo4j memory integration.
revise-claude-md
Update CLAUDE.md with learnings from this session.
check
View the latest screenshot from /Pictures/Screenshots.
deploy
Deploy all MCP plugin servers from .claude-plugin/marketplace.json.