Deep Research

What it does

When your agent uses it

• Any research/investigate/analyze request needing synthesis across web
• Community evidence — especially broad
• Ambiguous intent
• : research this

Example prompts

• “research this”
• “deep research”
• “investigate”
• “/deep-research”

Requirements

• Pre-approved tools (allowed-tools): Read, Grep, Glob, Bash, Write, WebSearch, WebFetch, Agent, Skill

Workflow steps

4 steps, taken from the step headings in SKILL.md.

1. 1Scope & Plan
2. 2Parallel Research
3. 3Synthesis + GapDetect
4. 4Conditional Validation

What it can do on your machine

Read from SKILL.md and the folder at commit c9a2036. It shows what the files ask for, not the result of running them.

• Tool permissions

  Pre-approves these tools, so the agent can use them without asking each time:

  • Read
  • Grep
  • Glob
  • Bash
  • Write
  • WebSearch
  • WebFetch
  • Agent
  • Skill

  From allowed-tools in the SKILL.md frontmatter.

• Runs code

  Shell commands in SKILL.md call:

  • git

  From the folder's file list and the shell code blocks in SKILL.md.

• Network

  No URLs in SKILL.md. Its commands use git, which can reach the network depending on how they are called.

  From URLs in SKILL.md, links to its own repository left out.

• Credentials

  Names no API keys, tokens, secrets or passwords.

  From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Deep Research loads about 3.5k tokens when it runs, and up to ~5.7k if it reads all its reference files. Until then it costs about 169 tokens; SKILL.md has 1,162 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it

~169

When it runs · the whole SKILL.md, loaded when a task matches

~3.5k

With references · SKILL.md plus every file in references/, read only if the agent opens them

~5.7k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

• NotePre-approves every shell command (allowed-tools: Bash)SKILL.md

  allowed-tools: Read, Grep, Glob, Bash, Write, WebSearch, WebFetch, Agent, Skill

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from sd0xdev/sd0x-harness at commit c9a2036, republished under its MIT licence (© sd0xdev). 1,162 words, ~3,537 tokens.

name

deep-research

description

Universal multi-source research orchestration. Use for any research/investigate/analyze request needing synthesis across web, codebase, and community evidence — especially broad, mixed, or ambiguous intent. Triggers on: 'research this', 'deep research', 'investigate', 'analyze from multiple angles', 'comprehensive analysis', 'explore this topic', 'study', 'survey the landscape', 'look into', 'understand deeply', '了解', '調查', '分析', '研究'. When intent is clearly single-dimension (code-only tracing, checklist-style compliance audit, or bounded option-ranking), dispatcher may prefer a narrower skill. Otherwise route here. Supports low/medium/high budget tiers.

allowed-tools

Read, Grep, Glob, Bash, Write, WebSearch, WebFetch, Agent, Skill

❌ git add | git commit | git push — per @rules/git-workflow.md

flowchart TD
    U[User: /deep-research topic] --> P0[Phase 0: Scope & Plan]
    P0 --> R[Phase 1: Parallel Research]
    R --> |2-3 agents| A1[Researcher: Web/Official]
    R --> |background| A2[Researcher: Code/Impl]
    R --> |background| A3[Researcher: Community/Cases]
    A1 --> S[Phase 2: Synthesis + GapDetect]
    A2 --> S
    A3 --> S
    S --> |claim registry| GATE{Score + Conflicts?}
    GATE --> |high score, no conflict| REPORT[Output Report]
    GATE --> |unresolved conflict or low score| V[Phase 3: Validation]
    V --> |validator micro-loop| VM[Dispute checks]
    VM --> |resolved| REPORT
    VM --> |still unresolved| DB[/codex-brainstorm]
    DB --> REPORT

## Research Plan: <topic>
- Intent: exploratory | compliance | decision
- Agents: N (shards: A=official, B=code, C=community)
- Budget: low | medium | high
- Scope: <path or "project root">

Agent({
  description: "Research shard A: <focus>",
  subagent_type: "Explore",  // or "general-purpose" as fallback
  run_in_background: true,
  prompt: <from references/research-roles.md researcher template>
})

## Deep Research Report: <topic>

### Research Metadata
- Mode: exploratory | compliance | decision
- Agents: N
- Sources: N (N official, N code, N community)
- Score: N/100 (confidence cap: X)

### Executive Summary
<synthesized answer to the research question>

### Findings by Source

| # | Claim | Evidence | Source Type | Confidence | Verified |
|---|-------|----------|------------|------------|----------|

### Claim Registry
| # | Claim | Sources | Consensus | Status |
|---|-------|---------|-----------|--------|

### Coverage Matrix
| Dimension | Score | Detail |
|-----------|-------|--------|
| Source diversity | N% | ... |
| Cross-verification | N% | ... |
| Gap coverage | N% | ... |
| Question closure | N% | ... |

### Divergence (if any)
| # | Claim A | Claim B | Resolution |
|---|---------|---------|------------|

### Debate Conclusion (if triggered)
- threadId: <from /codex-brainstorm>
- Rounds: N
- Equilibrium: <type>
- Key insight: <from debate>

### Residual Gaps & Next Steps
- <remaining unknowns>
- Suggested follow-up commands

Input: /deep-research "What are the best patterns for multi-agent orchestration?"
Output: 2-3 agents explore official docs + codebase + community → claim registry → score 85/100 → report with consensus findings

Input: /deep-research --mode compliance "Are our testing practices aligned with industry standards?"
Output: 3 agents → compliance mode forces debate → /codex-brainstorm equilibrium → gap analysis report

Input: /deep-research --mode decision "Should we use Redis or PostgreSQL for caching?"
Output: Parallel research on both options → claim registry with conflicts → debate on unresolved → recommendation with evidence

Input: /deep-research --budget low "What is WebAssembly?"
Output: Single inline research (no parallel agents) → lightweight report → score with 0.75 confidence cap

© sd0xdev, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 3 other files (references) in skills/deep-research of sd0xdev/sd0x-harness.

• SKILL.md
• references/claim-registry.md
• references/research-roles.md
• references/scoring-model.md

Open the folder on GitHubat commit c9a2036