improve-agent

A set of rules for improving an AI agent, which is software that performs tasks using language models and tools. It provides targeted recommendations for areas such as prompts, memory, security, performance, or verification.

In plain words
What is it for?
Use it when improving an agent’s orchestration, worker prompts, memory, tools, security, speed, or testing.
Why use it?
It gives focused improvement guidance based on the agent code and the selected area, instead of relying on general advice.

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/sarkarsaurabh27/agent-loop-learning/improve-agent
Clone the repo
git clone --depth 1 https://github.com/sarkarsaurabh27/agent-loop-learning

Made for: Cursor.

Per session 41 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 592 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 $0.00041 $0.00592
Opus 5 $0.00020 $0.00296
Sonnet 5 $0.00008 $0.00118
Haiku 4.5 $0.00004 $0.00059

Measured 2d ago against content hash 832e42deb20f, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

improve-agent 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.

.cursor/rules/improve-agent.mdc · 58 lines

What it actually says

Improve Agent Component

When the user asks to improve an agent or a specific component:

  1. Explore the current repo first — do not ask the user for anything yet. Search for agent-related code:
    • Files matching *agent*, *tool*, *prompt*, *chain*, *workflow* in src/, lib/, app/, root
    • Imports of agent frameworks: langchain, langgraph, openai, anthropic, autogen, crewai
    • System prompt definitions, tool definitions, agent loop logic
    • CLAUDE.md, AGENTS.md, or any README describing the agent architecture Read the relevant files. Only ask the user if no agent code is found after exploring.

Then map their keyword to the relevant best-practice doc:

Keyword Doc to read
orchestration / multi-agent best-practices/01-multi-agent-orchestration.md
worker / scaffold best-practices/02-worker-prompting.md
prompting / prompt best-practices/02-worker-prompting.md + 07-prompt-engineering.md
memory / context / rag best-practices/03-context-and-memory.md
tools / tool best-practices/04-tool-design.md
verification / testing / test best-practices/05-verification-and-testing.md
security / permissions / injection best-practices/06-security-and-permissions.md
performance / startup / latency best-practices/08-performance-and-startup.md
(no keyword / full system) all 9 docs

Read the matched doc(s), then produce improvement cards using this format:

### Improvement Plan: [component or "Full System"]

**Context** (1–2 sentences on current state)

---

**[Improvement title]**
- Current state: ...
- Recommended change: ...
- Benchmark justification: "[number] — [source, date]"
- Implementation sketch: (pseudocode or pattern, framework-agnostic)
- Effort: Low / Medium / High

---

#### Quick wins
1–3 changes under 1 hour with high impact.

Rules:

  • Prefer the simplest change that closes the gap.
  • Do not recommend adding frameworks unless directly needed.
  • If agent uses a specific model (GPT, Gemini, Llama), note model-specific vs. universal recommendations.
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 · 58 lines · 41 tokens per session scan A 832e42deb20f

Subscribe to this mod's changes

improve-agent is a cursor rule published in the GitHub repository sarkarsaurabh27/agent-loop-learning (3 stars, last pushed 3mo ago), licensed MIT. It adds 41 tokens to every session and 592 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.