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.
npx agentmods add rules/hlalljie/agent-workflow-presets/task-scopegit clone --depth 1 https://github.com/hlalljie/agent-workflow-presetsWhat 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 | $0.00476 | $0.00476 |
| Opus 5 | $0.00238 | $0.00238 |
| Sonnet 5 | $0.00095 | $0.00095 |
| Haiku 4.5 | $0.00048 | $0.00048 |
Grade A, and why
task-scope 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.
What it actually says
One task per message (default)
A task is one numbered block in a plan task file, e.g. ### Task N: … under .cursor/plans/**/tasks/*.md.
- Implement at most one such task per user message unless a skill the user invoked or a user instructions explicitly directs otherwise (e.g. run-task-loop — work through tasks in order in that session until done, blocked, or that skill's handoff).
- Multi-task only if the same message explicitly names more than one task number (e.g. "Task 3 and 5", "do 3, 5", "tasks 2–4").
- Or the task file's first line (the
# …title line) contains the literal substringbatchable: true. Then the user must still say they want batching in that message; the flag only allows it, it does not auto-batch.
Forbidden expansions
Without a new message that names another task number, do not:
- Implement the next task because tests, MSW, build, or "verification" need it.
- Add "small" wrappers, clients, or handlers for a different task section.
- Mark additional checkboxes complete in the plan for work you were not asked to do.
If work truly depends on another task, state the dependency in prose and stop. Do not implement the dependency unless the user then names that task.
Discussion is not implementation
If the user is venting, reviewing a mistake, asking why, or debating process: no repo edits unless they give a separate, explicit implement instruction (e.g. "apply that patch", "revert task 4").
Ordering
If the user references a task file but no task number: do not guess. Ask which Task N to run (one line).
To author a new task file from a plan phase, use the create-task-list skill.
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.
- 2d ago First seen · 33 lines · 476 tokens per session scan A 624e8e5ff4a3
task-scope is a cursor rule published in the GitHub repository hlalljie/agent-workflow-presets (2 stars, last pushed 1mo ago), licensed MIT. It adds 476 tokens to every session, about $0.0024 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.
Other cursor rules, from other repositories
honey
Honey — write less code and say less about it. A reflexive minimal-code (YAGNI/stdlib-first) and terse-prose discipline that cuts token cost while keeping code, commands, and safety-critical paths exact.
module-shape
Prefer modules with a narrow public surface and a thick internal implementation. Inject external dependencies at the boundary. Mock only at system edges.
context-discipline
Manage AI context window efficiently - read before write, no re-reads, tool-call budgets, one-pass discipline.
quality-gates
Enforce quality checks before commits - lint, typecheck, and test affected code.
token-efficiency
Token efficiency rules - eliminate waste in AI output, enforce read-before-write, prevent iteration cycles.
pn-nextjs
Next.js best practices. Data loading, server/client boundaries, streaming, mutations, and performance. For Next app/ or pages/, pn-react also applies (core React patterns); content is complementary.