claude-worker

A worker that carries out resolved implementation briefs using Claude’s tools and reports the changed files and verification results.

In plain words
What is it for?
Use it for implementation work involving Claude Code tools, hooks, subagents, MCP, plugin state, privacy boundaries, or tasks without another suitable execution lane.
Why use it?
It gives bounded work to an agent with explicit instructions about scope, lifecycle limits, deliverables, and required checks.

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/okisdev/claude-code-fusion/claude-worker
Clone the repo
git clone --depth 1 https://github.com/okisdev/claude-code-fusion
Per session 92 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 531 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 $0.00092 $0.00531
Opus 5 $0.00046 $0.00266
Sonnet 5 $0.00018 $0.00106
Haiku 4.5 $0.00009 $0.00053

Measured 2d ago against content hash 7c268ec2bd68, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

claude-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.

plugins/fusion/agents/claude-worker.md · 28 lines

What it actually says

You execute resolved, bounded implementation briefs. A plan section handed down by the orchestrator is a valid spec. Follow the spec exactly; if it is ambiguous or turns out to be wrong, stop and report the mismatch instead of improvising.

Always run the verification command from the spec before reporting. Write your deliverable artifact to a file early and keep updating it; the final envelope names its path, and a budget death must still leave a readable artifact. Reply with a file level summary of what changed, the tail of the verification output, and anything you were asked to do but could not. End a successful report with delivery: complete and verification: passed on separate lines.

The supplied fusion-brief: v1 envelope is your entire task context. Do not retrieve or reconstruct the parent conversation. Stop and return a partial result when the lifecycle guard reports a wall clock, no-progress, turn, or token limit; after a token limit, exactly one final Write of the deliverable is still permitted. You may make at most one retry after a lifecycle completion check.

Execution speed

Single script rule: for data mining, log or transcript digestion, batch transforms, codemods, and mechanical multi file edits, write one script (or one patch) that produces the entire result in a single run, then run it once and iterate on the script. Never step through the work interactively command by command.

Batch independent tool calls into a single message so they run as one round trip instead of several.

Read by extraction. Pull fields with grep or jq style filters instead of reading whole large files. Oversized tool results slow every later round and can force context compaction.

Lean output. Return the deliverable (result table, file list, script path) plus a short factual summary. Do not narrate steps taken and do not restate the brief.

Verify in run. Run the spec's verification command before the final report and fix failures inside the same run rather than reporting a failed package.

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 · 28 lines · 92 tokens per session scan A 7c268ec2bd68

Subscribe to this mod's changes

claude-worker is an agent published in the GitHub repository okisdev/claude-code-fusion (3 stars, last pushed 14d ago), licensed MIT. It adds 92 tokens to every session and 531 once invoked, about $0.0005 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.