context-management

Rules for keeping track of work during long coding sessions, including splitting large changes into smaller tasks and summarising progress. A context window is the amount of conversation and code the agent can consider at once.

In plain words
What is it for?
Use it for multi-step features, sessions lasting several turns, and code discussions that require full project file paths.
Why use it?
It reduces the chance that an agent forgets earlier decisions, loses track of unfinished work, or confuses similarly named files.

Cursor rule for Cursor

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 rules/vibestackdev/vibe-stack/context-management
Clone the repo
git clone --depth 1 https://github.com/vibestackdev/vibe-stack

Made for: Cursor.

Per session 0 Nothing until a file matches its globs; then the whole rule loads.
When invoked 454 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 $0.00000 $0.00454
Opus 5 $0.00000 $0.00227
Sonnet 5 $0.00000 $0.00091
Haiku 4.5 $0.00000 $0.00045

Measured yesterday against content hash 77cd9736ab59, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

context-management 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 yesterday.

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.

.cursor/rules/context-management.mdc · 62 lines

What it actually says

Context Management Rules

RULE 1: Break Large Tasks Into Sub-Tasks

If a feature requires changes to more than 4 files, break it into sequential sub-tasks. Complete and verify each sub-task before starting the next.

Example — "Add a blog feature with CRUD":

  1. Sub-task 1: Database migration + RLS policies
  2. Sub-task 2: Server Actions (create, read, update, delete)
  3. Sub-task 3: UI components (list, detail, form)
  4. Sub-task 4: Route handlers + API endpoints
  5. Sub-task 5: Integration testing

RULE 2: Summarize Before Continuing

If the conversation has been going for more than 5 turns, summarize what has been done and what remains before generating more code.

Template:

Before continuing, let me summarize:
✅ Completed: [list of completed items]
🔄 In progress: [current item]
📋 Remaining: [list of remaining items]

RULE 3: Reference Files By Full Path

When discussing code, ALWAYS use the full path relative to the project root. This prevents confusion between similarly named files.

✅ CORRECT: "Update src/app/dashboard/page.tsx" ❌ WRONG: "Update the dashboard page"

✅ CORRECT: "Import from @/lib/supabase/server" ❌ WRONG: "Import the Supabase client"

RULE 4: One Concern Per Message

Each AI message should address ONE concern clearly:

  • ONE file modification with before/after
  • ONE bug fix with explanation
  • ONE new feature with complete implementation

Do NOT mix unrelated changes in a single response. This makes review harder and increases the chance of introducing bugs.

RULE 5: Keep Rule Files Under 150 Lines

Individual .mdc rule files should stay under 150 lines. If a rule file grows beyond that, split it into focused sub-rules.

This mirrors the principle from production AI harnesses: keep the index lightweight, put details in separate files.

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. yesterday First seen · 62 lines · 0 tokens per session scan A 77cd9736ab59

Subscribe to this mod's changes

context-management is a cursor rule published in the GitHub repository vibestackdev/vibe-stack (7 stars, last pushed 2d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 454 tokens. 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.