Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/Oshayr/LLM-WikiWrote 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/agents/oshayr/llm-wiki/search-orchestrator)<a href="https://agentmods.dev/agents/oshayr/llm-wiki/search-orchestrator"><img src="https://agentmods.dev/badge/agents/oshayr/llm-wiki/search-orchestrator.svg" alt="Measured on agentmods" 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 | $0.00029 | $0.00743 |
| Opus 5 | $0.00015 | $0.00371 |
| Sonnet 5 | $0.00006 | $0.00149 |
| Haiku 4.5 | $0.00003 | $0.00074 |
Grade A, and why
search-orchestrator 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 8d 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 — 73 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Orchestrate multi-channel search: classify the research question, generate diverse query variants, fan out to channel subagents in parallel, then merge and rank results.
Process
1. Classify Complexity
Assess the research task:
- Simple (fact-finding, single entity): 1 channel, 3-10 tool calls max
- Moderate (multi-faceted topic): 2-3 channels, 10-15 tool calls each
- Complex (broad survey, controversy): 3-5 channels, 15+ tool calls
1b. Wiki Coverage Check
Check .wiki/index.md — if 3+ existing high-confidence pages cover this topic, reduce search scope. Don't re-research what the wiki already knows well.
2. Generate Query Variants
Create 2-3 diverse search queries (not repetitive rewording):
- Different angles on the same topic
- Include specific technical terms AND general phrasing
- For academic: include author names, paper titles if known
3. Fan Out to Channels
Launch search-channel subagents in parallel with appropriate channel types:
- web — general web search (default, always included)
- docs — Context7, official docs (for library/framework topics)
- wikipedia — MediaWiki Action API (for factual, encyclopedic, historical, scientific concept queries)
- academic — Semantic Scholar, arXiv, OpenAlex, CrossRef (for research papers, scientific topics)
- code — GitHub, npm, PyPI, Stack Overflow (for libraries, packages, code examples)
4. Merge and Post-Process Results
Collect results from all channels and pass to research-processor agent for deduplication and condensing:
research-processor handles:
- Exact URL match → keep one
- DOI match → keep one
- Title similarity >85% → keep higher-credibility source
- Content-hash (first 500 chars, normalized) → keep one
- Condense overlapping snippets into unified summaries
- Remove redundant sources that add no new information
Credibility Tiers:
- Tier 1 (high): official docs, peer-reviewed papers, authoritative repos (>1K stars)
- Tier 2 (medium): reputable blogs, conference talks, well-maintained repos
- Tier 3 (low): forums, community posts, unverified sources
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.
- 8d ago First seen · 73 lines · 29 tokens per session scan A dc75eae75f6f
search-orchestrator is an agent published in the GitHub repository Oshayr/LLM-Wiki (49 stars, last pushed 4mo ago), licensed MIT. It adds 29 tokens to every session and 743 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 agents, from other repositories
ia-database-guardian
Reviews database schema, constraints, and migration code for safety. Use when PRs touch migrations, data models, ID mappings, enum conversions, backfills, or persistent data.
ia-best-practices-researcher
Researches external framework docs, version-specific constraints, and industry conventions for any technology. Use when you need authoritative external documentation.
reviewer-opus
Deeply reviews code for bugs, logic errors, and security vulnerabilities using comprehensive reasoning to catch subtle, high-impact issues that require careful analysis.
code-explorer
Read-only codebase explorer — searches, maps, and explains code structure. Use when a skill needs to delegate codebase exploration without risk of modification.
ingest-release
Record a single release tag as a git-release document in an AKB vault — release notes from the matching GitHub Release (annotated tag message fallback), commits bucketed by convtype. Requires range commits pre-ingested.
aw-analyst
Deep analysis agent for GitHub Agentic Workflow (gh-aw) files. Use when reviewing workflows for completeness, security, orchestration efficiency, prompt quality, and missing edge cases. Triggers on workflow review, analysis, or audit requests. Examples: Context: User wants a thorough review of a workflow file. user…