Getting it into your agent
This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.
/plugin marketplace add rvk7895/llm-knowledge-bases/plugin install kbWrote 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/skills/rvk7895/llm-knowledge-bases/research-deep)<a href="https://agentmods.dev/skills/rvk7895/llm-knowledge-bases/research-deep"><img src="https://agentmods.dev/badge/skills/rvk7895/llm-knowledge-bases/research-deep/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/skills/rvk7895/llm-knowledge-bases/research-deep"><img src="https://agentmods.dev/badge/skills/rvk7895/llm-knowledge-bases/research-deep.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 | $0.00019 | $0.00887 |
| Opus 5 | $0.00010 | $0.00443 |
| Sonnet 5 | $0.00004 | $0.00177 |
| Haiku 4.5 | $0.00002 | $0.00089 |
Grade A, and why
research-deep 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 9d 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 — 100 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Attribution: Originally authored by Weizhena. Included with attribution for use in the Deep query workflow.
Research Deep - Deep Research
Trigger Method
/research-deep
Execution Flow
Step 1: Auto-locate Outline
Find */outline.yaml file in current working directory, read items list and execution configuration (including items_per_agent).
Step 2: Resume From Checkpoint
- Check completed JSON files in output_dir
- Skip already completed items
Step 3: Batch Execution
- Execute in batches by batch_size (need user approval to proceed to next batch after completing one)
- Each agent handles items_per_agent projects
- Launch web-search-agent (background parallel, task output disabled)
Parameter Retrieval:
{topic}: topic field from outline.yaml{item_name}: name field of item{item_related_info}: complete yaml content of item (name + category + description etc.){output_dir}: execution.output_dir from outline.yaml (default ./results){fields_path}: absolute path to {topic}/fields.yaml{output_path}: absolute path to {output_dir}/{item_name_slug}.json (slugify item_name: replace spaces with _, remove special characters)
Hard Constraint: The following prompt must be strictly recited, only replace variables in {xxx}, do not rewrite structure or wording.
Prompt Template:
prompt = f"""## Task
Research {item_related_info}, output structured JSON to {output_path}
## Field Definitions
Read {fields_path} to get all field definitions
## Output Requirements
1. Output JSON according to fields defined in fields.yaml
2. Mark uncertain field values as [不确定]
3. Add uncertain array at end of JSON, listing all uncertain field names
4. All field values must be output in Chinese (research process can use English, but final JSON values in Chinese)
## Output Path
{output_path}
## Validation
After completing JSON output, run validation script to ensure complete field coverage:
python ~/.claude/skills/research/validate_json.py -f {fields_path} -j {output_path}
Task is only complete after validation passes.
"""
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.
- 9d ago First seen · 100 lines · 19 tokens per session scan A 3770a5df327c
research-deep is a skill published in the GitHub repository rvk7895/llm-knowledge-bases (37 stars, last pushed 2mo ago), licensed MIT. It adds 19 tokens to every session and 887 once invoked, about $0.0001 per session on Opus 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 skills, from other repositories
link-memory
Use after important user-approved decisions, when durable context should be proposed or reviewed, and for explicit Link memory lifecycle work: remember, recall, review, update, archive, restore, forget, or explain local memories through the CLI without requiring MCP.
link-retrieve
Use before answering work that may depend on user memory, project history, source-backed notes, or prior decisions; retrieve compact Link context through the CLI without loading the whole wiki or requiring MCP.
link-health
Use at the start of Link work when readiness is unclear, after installs or upgrades, and before repairs; verify health, inspect interrupted writes, back up, and repair generated indexes without MCP.
link-ingest
Use when raw files are present, source pages look stale, or a user asks to ingest notes into Link; refresh source-backed wiki pages, propose memories, and validate updates through the CLI without MCP.
superbrain-distill
Internal SuperBrain skill — run by the detached capture child to distill a session-event delta into routed Obsidian notes. Not for direct user invocation.
superbrain-recall
Search the user's SuperBrain second-brain vault. Use whenever the user references past work, prior decisions, "how did we", "did we already", earlier sessions, a project's history, or anything that may already be recorded — before answering from scratch.