delegate-work

A guide for handing one independent, well-defined task to a separate analyst or reviewer that works asynchronously.

In plain words
What is it for?
Use it to request focused data inspection, experiments, generated artifacts, or read-only reviews with clear acceptance criteria.
Why use it?
It keeps the main work moving while a separate agent handles a task that needs substantial context. It also prevents vague or overly broad delegations.

Skill for Claude CodeCodex

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 skills/aws-samples/sample-strands-agent-with-agentcore/delegate-work
Any agent
npx skills add aws-samples/sample-strands-agent-with-agentcore --skill delegate-work
Clone the repo
git clone --depth 1 https://github.com/aws-samples/sample-strands-agent-with-agentcore

Made for: Claude Code, Codex.

Per session 58 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 492 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 $0.00058 $0.00492
Opus 5 $0.00029 $0.00246
Sonnet 5 $0.00012 $0.00098
Haiku 4.5 $0.00006 $0.00049

Measured 3d ago against content hash 41571c142df3, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

delegate-work 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 3d 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.

chatbot-app/agentcore/skills/delegate-work/SKILL.md · 58 lines

How it starts

The opening of the file, as written. The whole thing — 58 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Delegate Work

Delegate only work that is independent enough to finish in an isolated context.

Scope the task

  • Give one concrete goal and one deliverable.
  • Provide observable acceptance criteria.
  • Pass only the workspace paths and context required for the task.
  • State explicit non-goals in constraints.
  • Never copy the full conversation into context_summary.
  • Split unrelated deliverables into separate delegations.

Do not delegate requests such as "investigate everything", "fix all issues", or "continue helping with this project". Narrow them first.

Select a profile

  • Use analyst for data inspection, computation, experiments, and generated artifacts. It can use the session Code Interpreter.
  • Use reviewer for an independent read-only review. It cannot execute code or modify source files.

Select task complexity

  • Use low for narrow extraction, classification, or simple checks.
  • Use medium for normal analysis and review work.
  • Use high only for genuinely difficult reasoning or broad synthesis.
  • Omit task_complexity when the delegated agent should inherit the parent model unchanged.

Execute

Call delegate_task once. It returns an acceptance receipt immediately. Continue the foreground conversation without waiting for completion. Never expose or invent internal job identifiers; they are orchestration metadata owned by the activity UI and backend.

Use get_delegation only when the user asks for status or the result is needed before another decision. Filter by profile or distinctive goal text. Use cancel_delegation only for an explicit request to cancel a background job; if multiple jobs match, narrow the goal instead of guessing.

Boundaries

  • Do not delegate recursively.
  • Do not use delegation for Gmail, Calendar, Notion, GitHub writes, OAuth, or other external side effects.
  • Do not ask the subagent to modify files outside its output directory.
  • Do not assume an interrupt of the foreground response cancels an accepted delegation.
  • Treat the mailbox completion as the authoritative terminal result.

Read the full file on GitHub · 58 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. 3d ago First seen · 58 lines · 58 tokens per session scan A 41571c142df3

Subscribe to this mod's changes

delegate-work is a skill published in the GitHub repository aws-samples/sample-strands-agent-with-agentcore (191 stars, last pushed 7d ago), licensed MIT. It adds 58 tokens to every session and 492 once invoked, about $0.0003 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 2026-08-30.