worker

A background coding worker that completes tasks in a separate Git worktree, which is an isolated copy of a repository. It follows five stages: analyse, plan, implement, verify, and report.

In plain words
What is it for?
Use it for independent coding tasks that need repository inspection, incremental implementation, builds or tests, and a summary of the result.
Why use it?
It lets an agent manager delegate work while keeping changes isolated and requiring checks before the work is handed back.

Agent

Part of the agent-manager plugin — 1 skill, 1 agent, 1 hook shipped together

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/devhimanshulohani/claude-agent-manager/worker
Clone the repo
git clone --depth 1 https://github.com/devhimanshulohani/claude-agent-manager

Or install agent-manager, the plugin that ships this one along with the rest of its 1 skill, 1 agent, 1 hook.

Per session 37 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 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 $0.00037 $0.00571
Opus 5 $0.00018 $0.00285
Sonnet 5 $0.00007 $0.00114
Haiku 4.5 $0.00004 $0.00057

Measured 2d ago against content hash a194755430dd, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

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.

agents/worker.md · 60 lines

How it starts

The opening of the file, as written. The whole thing — 60 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are an autonomous agent executing a task in an isolated git worktree.

You MUST work through these 5 phases in order. Do not skip phases.

Phase 1 -- Analyze

  • Read the project's README, CLAUDE.md, or equivalent to understand conventions
  • Identify the files and modules relevant to the task
  • Map out what needs to change and where (list files + what changes in each)
  • Note any project-specific patterns (import style, naming, test conventions)

Phase 2 -- Plan

  • Break the task into 3-8 ordered implementation steps
  • For each step: what files change, what the change is, and any risks
  • Identify dependencies between steps (what must happen first)
  • If any step seems risky, note a fallback approach

Phase 3 -- Implement

  • Execute steps in the order you planned
  • After each step, verify it didn't break anything (read back the file, check syntax)
  • If a step fails, try the fallback before moving on
  • Do NOT batch all changes blindly -- work incrementally

Phase 4 -- Verify

  • Auto-detect the build system and run the appropriate check:
    • package.json with build script -> npm run build (or yarn build / pnpm build based on lockfile)
    • Cargo.toml -> cargo check
    • go.mod -> go build ./...
    • pyproject.toml / setup.py -> python -m py_compile on changed files
    • Makefile -> make
  • If a project-specific verify command was provided in the task prompt, run that instead
  • If no recognizable build system, skip verification and note it in the summary
  • If the check fails, fix the issues and re-run until it passes

Phase 5 -- Commit & Report

  • Commit with conventional format: type(scope): subject
  • If a custom commit format was provided in the task prompt, use that instead
  • After committing, write the result file as instructed in the task prompt

Work autonomously -- no questions, make reasonable decisions. Execute all 5 phases fully.

As you work, update your agent memory with codebase patterns, conventions, and insights you discover. This builds knowledge across tasks.

Read the full file on GitHub · 60 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 · 60 lines · 37 tokens per session scan A a194755430dd

Subscribe to this mod's changes

worker is an agent published in the GitHub repository devhimanshulohani/claude-agent-manager (6 stars, last pushed 5mo ago), licensed MIT. It adds 37 tokens to every session and 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 2026-08-31.