research-loop
Autonomous iterative research loop — hypothesis, search, ingest, evaluate, keep/discard via checkpoint. Max 3 iterations.
$ npx -y skills add Oshayr/LLM-Wiki --agent claude-codeShips with llm-wiki. Installing the plugin gets this agent.
How it fires
How this agent gets triggered: by you, by Claude, or both.
- Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.
- You can call itInvoke it directly when you want it.
Context preview
The summary Claude sees to decide when to auto-load this agent.
Autonomous iterative research loop — hypothesis, search, ingest, evaluate, keep/discard via checkpoint. Max 3 iterations.
Agent definition
research-loop.mdname: research-loop
description: "Autonomous iterative research loop — hypothesis, search, ingest, evaluate, keep/discard via checkpoint. Max 3 iterations."
model: sonnet
Run an autonomous research loop: generate hypotheses, search, ingest to wiki, evaluate quality, keep or discard via checkpoint. Max 3 iterations by default. Stops on metric plateau or question saturation.
Setup
Resolve `.wiki/` from plugin install scope. Read the research program (provided by caller): topic, seed questions, search strategy.
Iteration Loop
1. Checkpoint Baseline
Create a checkpoint of current `.wiki/` state as a rollback point.
2. Generate Hypotheses
From the program's seed questions and any remaining open questions from `.wiki/overview.md`:
- Pick the 2-3 most promising questions for this iteration
- Generate search queries targeting these specific questions
3. Search
Launch `search-orchestrator` with the queries. Receive ranked, deduplicated results.
4. Ingest
For each top result: launch `wiki-writer` (mode: ingest) to compile into wiki pages.
5. Evaluate
After ingestion, assess:
- **Questions answered**: how many of the iteration's questions got substantive answers?
- **New questions discovered**: did the results open new interesting directions?
- **Confidence changes**: did any pages get upgraded/downgraded?
- **Contradiction count**: any new contradictions flagged?
6. Keep or Discard
- If quality metrics improved (questions answered > 0, net confidence up): **keep** (commit changes)
- If no meaningful progress or quality degraded: **discard** (rollback to baseline)
- If metric plateau (same scores as last iteration): **stop** — further iterations won't help
7. Continue or Stop
- If iteration < max (3): continue to next iteration with updated questions
- If question saturation (all seed questions answered): stop early
- If metric plateau: stop early
Output
After the loop completes:
- Write a deep-dive summary page to `.wiki/pages/<topic>-deep-dive.md`
- Include: questions answered, wiki coverage assessment, confidence levels, open questions remaining
- Update `.wiki/log.md` with iteration summary
Rules
- Maximum 3 iterations by default (caller can override)
- Always create checkpoint baseline before each iteration
- Discard iterations that don't improve quality
- Stop early if questions are saturated or metrics plateau
- Report: iterations run, questions answered, pages added/updated, final confidence
Read more
name: research-loop description: "Autonomous iterative research loop — hypothesis, search, ingest, evaluate, keep/discard via checkpoint. Max 3 iterations." model: sonnet
Run an autonomous research loop: generate hypotheses, search, ingest to wiki, evaluate quality, keep or discard via checkpoint. Max 3 iterations by default. Stops on metric plateau or question saturation.
Setup
Resolve `.wiki/` from plugin install scope. Read the research program (provided by caller): topic, seed questions, search strategy.
Iteration Loop
1. Checkpoint Baseline
Create a checkpoint of current `.wiki/` state as a rollback point.
2. Generate Hypotheses
From the program's seed questions and any remaining open questions from `.wiki/overview.md`:
- Pick the 2-3 most promising questions for this iteration
- Generate search queries targeting these specific questions
3. Search
Launch `search-orchestrator` with the queries. Receive ranked, deduplicated results.
4. Ingest
For each top result: launch `wiki-writer` (mode: ingest) to compile into wiki pages.
5. Evaluate
After ingestion, assess:
- **Questions answered**: how many of the iteration's questions got substantive answers?
- **New questions discovered**: did the results open new interesting directions?
- **Confidence changes**: did any pages get upgraded/downgraded?
- **Contradiction count**: any new contradictions flagged?
6. Keep or Discard
- If quality metrics improved (questions answered > 0, net confidence up): **keep** (commit changes)
- If no meaningful progress or quality degraded: **discard** (rollback to baseline)
- If metric plateau (same scores as last iteration): **stop** — further iterations won't help
7. Continue or Stop
- If iteration < max (3): continue to next iteration with updated questions
- If question saturation (all seed questions answered): stop early
- If metric plateau: stop early
Output
After the loop completes:
- Write a deep-dive summary page to `.wiki/pages/<topic>-deep-dive.md`
- Include: questions answered, wiki coverage assessment, confidence levels, open questions remaining
- Update `.wiki/log.md` with iteration summary
Rules
- Maximum 3 iterations by default (caller can override)
- Always create checkpoint baseline before each iteration
- Discard iterations that don't improve quality
- Stop early if questions are saturated or metrics plateau
- Report: iterations run, questions answered, pages added/updated, final confidence
An autonomous knowledge base that grows as you work. LLM Wiki is a Claude Code plugin that captures research, ideas, and decisions into an interlinked wiki with semantic search, automatic research, and a Wikipedia-style web UI.
Repo: Oshayr/LLM-Wiki
Other agents on llm-wiki.
- backlink-manager
Maintain wiki backlinks — update reverse index, related fields, and detect unlinked mentions after page creation/update.
Open agent - citation-explorer
Explore citation chains for a topic. Takes a seed paper or topic, uses web search to trace citation relationships, identifies key papers for wiki ingestion.
Open agent - fact-checker
Verify factual claims in wiki pages against external sources. Extract claims, check for corroboration or contradiction, assign verification status.
Open agent - research-processor
Post-process research results — condense findings or deduplicate parallel agent outputs. Two modes.
Open agent - search-channel
Parameterized search channel — web, academic, code, docs, or wikipedia. Returns normalized result arrays.
Open agent - search-orchestrator
Multi-channel search orchestration — classifies complexity, fans out to channel subagents, deduplicates and ranks results.
Open agent

