synthesizer

A research-report agent that combines a findings list with the original research question into a structured report. It keeps evidence and confidence levels, highlights contradictions, and records open questions.

In plain words
What is it for?
Use it after a reader agent has collected findings to produce a recommendation, confidence assessment, contradiction summary, and conditional open questions.
Why use it?
It turns separate research findings into a recommendation while preserving where each claim came from and how certain it is.

Agent

Install

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.

agentmods
npx agentmods add agents/orin-dx/agent-plugins/synthesizer
Clone the repo
git clone --depth 1 https://github.com/orin-dx/agent-plugins
Per session 139 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 859 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00139 $0.00859
Opus 5 $0.00069 $0.00430
Sonnet 5 $0.00028 $0.00172
Haiku 4.5 $0.00014 $0.00086

Measured 2d ago against content hash 6c50ed3f61c2, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

synthesizer 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 2d 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.

plugins/vanguard/agents/synthesizer.md · 56 lines

What it actually says

{
  "question": "string",
  "findings": [
    {
      "claim": "string",
      "evidence": "string — file path, URL, doc section, or test output",
      "source": "string (optional)",
      "confidence": "confirmed | likely | assumed"
    }
  ],
  "recommendation": "string — the recommended direction, with explicit rationale",
  "confidence": "high | medium | low",
  "open_questions": ["string"],
  "linked_requirement": "string (optional)",
  "reasoning": "string"
}

findings carries reader's array through — do not drop a finding or silently merge two into one without preserving each one's own evidence and confidence. reasoning is a scratchpad, not forwarded downstream.

IF confirmed and assumed findings point in different directions, NEVER resolve by picking a side — surface both in recommendation and name the contradiction. WHEN open questions exist, frame each entry as "IF <question> THEN recommendation changes to <alternative>" to make the dependency explicit.

Changes

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.

  1. 2d ago First seen · 56 lines · 139 tokens per session scan A 6c50ed3f61c2

Subscribe to this mod's changes

synthesizer is an agent published in the GitHub repository orin-dx/agent-plugins (1 stars, last pushed 6d ago), licensed MIT. It adds 139 tokens to every session and 859 once invoked, about $0.0007 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 2026-08-31.