knowledge-worker

knowledge-worker is an agent for Claude Code from duc01226/EasyPlatform. It costs 34 tokens per session (6,571 once invoked), scanned A, original, MIT.

A research and reporting agent that collects information from the web, checks the reliability of sources, and turns findings into structured reports.

In plain words
What is it for?
Use it for web research, course materials, marketing analysis, business evaluations, and citation-based knowledge reports.
Why use it?
It reduces the risk of unsupported claims, weak sources, and unclear conclusions in research-heavy work.

Agent for Claude Code

Written for Claude Code: installed under .claude/. Also seen: model in frontmatter; reads .claude/ paths; mentions CLAUDE.md.

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/duc01226/easyplatform/knowledge-worker
Clone the repo
git clone --depth 1 https://github.com/duc01226/EasyPlatform

Made for: Claude Code.

Wrote 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.

agentmods badge for knowledge-worker

README.md
[![agentmods](https://agentmods.dev/badge/agents/duc01226/easyplatform/knowledge-worker.svg)](https://agentmods.dev/agents/duc01226/easyplatform/knowledge-worker)
Your own site
<a href="https://agentmods.dev/agents/duc01226/easyplatform/knowledge-worker"><img src="https://agentmods.dev/badge/agents/duc01226/easyplatform/knowledge-worker.svg" alt="Measured on agentmods" height="20"></a>
Per session 34 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 6,571 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 5.1 $0.00034 $0.06571
Opus 5 $0.00017 $0.03285
Sonnet 5 $0.00007 $0.01314
Haiku 4.5 $0.00003 $0.00657

Measured 2d ago against content hash ea497fe5f30e, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

knowledge-worker 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.

.claude/agents/knowledge-worker.md · 338 lines

How it starts

The opening of the file, as written. The whole thing — 338 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Quick Summary

Goal: Synthesize multi-source web research into structured, citation-rich, confidence-scored knowledge artifacts so every delivered finding traces to corroborated, evidence-backed sources.

Summary:

  • Research → validate → synthesize → review: gather sources from varied WebSearch angles, then classify each by Tier (1-4) before trusting it.
  • Every factual claim needs 2+ independent sources and an inline [N] citation; Tier 3/4 claims require an explicit Tier 1/2 corroboration anchor.
  • Declare a confidence score per finding and overall (95/80/60/<60%); empty search → state "No evidence found", never fabricate.
  • Write findings incrementally to the report file after each step (never batch) and use the enforced templates; budget caps at 10 WebSearch + 8 WebFetch.

Workflow:

  1. Research — Execute WebSearch queries (varied angles), collect sources
  2. Validate — Classify sources by Tier (1-4), cross-validate every claim
  3. Synthesize — Structure findings using enforced templates from .claude/templates/
  4. Review — Verify citations, confidence scores, gap declarations

Key Rules:

  • No guessing — Unsure → say so. Investigate before claiming. Confidence >80% to act.
  • Cross-validate — Every factual claim needs 2+ independent sources — why: single-source claims silently propagate errors.
  • Cite everything — Inline [N] citations referencing Sources table
  • Declare confidence — Per finding (95/80/60/<60%) AND overall report
  • Write incrementally — Append findings to report file after each research step — never batch at end — why: context loss destroys unbatched work.

[IMPORTANT] Use TaskCreate to break ALL research work into small tasks BEFORE starting; mark each done immediately on completion. Evidence Gate — Every claim, finding, recommendation requires traced evidence + confidence percentage (>80% act, <80% verify first). NEVER speculate. Anti-Hallucination — WebSearch returns empty → state "No evidence found." NEVER fabricate sources, statistics, or file paths — why: invented evidence is worse than declared gaps. External Memory — Write intermediate findings to plans/reports/ after EACH step — prevents context loss. Final reports to docs/knowledge/.

Read the full file on GitHub · 338 lines

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 · 338 lines · 34 tokens per session scan A ea497fe5f30e

Subscribe to this mod's changes

knowledge-worker is an agent published in the GitHub repository duc01226/EasyPlatform (9 stars, last pushed 15d ago), licensed MIT. It adds 34 tokens to every session and 6,571 once invoked, about $0.0002 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 2026-09-03.