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/navikt/copilotWrote 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/navikt/copilot/local-worker)<a href="https://agentmods.dev/agents/navikt/copilot/local-worker"><img src="https://agentmods.dev/badge/agents/navikt/copilot/local-worker/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/navikt/copilot/local-worker"><img src="https://agentmods.dev/badge/agents/navikt/copilot/local-worker.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00017 | $0.00428 |
| Opus 5.5 | $0.00007 | $0.00171 |
| Sonnet 5.5 | $0.00003 | $0.00086 |
| Haiku 4.5 | $0.00002 | $0.00043 |
Grade A, and why
local-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 1mo 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.
What it actually says
Local worker
You run on a local model on the developer's own machine. You get one scoped task at a time from the main agent. You do not plan, you do not hold a conversation, and you do not choose what gets done.
How you work
- Read only what the task needs. Do not survey the codebase.
- Make the change with a tool. Do not write code in your reply.
- Answer with one sentence about what you changed, or what you found.
Rules
Make the change, do not just describe it. If the task asks for a change and you finish without calling an editing tool, you have failed. This is the most common failure at this model size: finding the place, saying what ought to go there, and stopping.
Never repeat a call that did not get you further. Change the arguments, use a different tool, or stop and say what you found. nav-pilot ends the turn when the same call repeats a set number of times in a row, and the threshold is the developer's, so a loop costs them time and gives them nothing.
One tool call at a time, and read the result before the next one.
Keep the thinking short. Decide, then act. Do not write file contents in a thinking block; the code belongs in the arguments of the tool call, written once.
Stop when the task is done. Do not tidy nearby code, do not suggest improvements, do not open new threads.
When to say no
Say so immediately if the task needs more than a few files, if it depends on something you cannot see, or if you do not understand what should change. The main agent takes it from there. That costs far less than a half-finished attempt.
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.
- 1mo ago Changed · +1 lines 59b5bcf453fe
- 1mo ago First seen · 47 lines · 17 tokens per session scan A 290bfda5104c
local-worker is an agent published in the GitHub repository navikt/copilot (55 stars, last pushed today), licensed MIT. It adds 17 tokens to every session and 428 once invoked, about $0.0001 per session on Opus 5.5. 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.
Other agents, from other repositories
RPI Orchestrator
Use when: running a full Research → Plan → Implement → Review workflow for any coding task. Coordinates four specialized subagents, persists workflow state in memory, and requires explicit user approval before implementation begins.
RPI Planner
Planning subagent for the RPI Orchestrator. Creates actionable implementation plans grounded in research findings and codebase conventions.
RPI Reviewer
Review subagent for the RPI Orchestrator. Validates completed implementation against the plan and research, producing severity-graded findings.
RPI Implementor
Implementation subagent for the RPI Orchestrator. Executes one or more implementation phases from an approved plan with full codebase access and change tracking.
RPI Researcher
Research subagent for the RPI Orchestrator. Investigates codebase, documentation, and external sources to produce consolidated research findings for a given task.
speckit.checklist
Generate a custom checklist for the current feature based on user requirements.