agent-memory

A rule for keeping persistent notes about an AI agent's past mistakes, fixes, and useful project knowledge in an agent-memory file.

In plain words
What is it for?
Use it on every request to read the stored memory and, when a new mistake or better fix is found, record the correction for later work.
Why use it?
It helps the agent avoid repeating known errors across requests by consulting previous corrections before working.

Cursor rule

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/e-gov/cursor-prompts/agent-memory
Clone the repo
git clone --depth 1 https://github.com/e-gov/cursor-prompts
Per session 450 This file is loaded in full into every session.
When invoked 450 The same file — it is already loaded in full.
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 $0.00450 $0.00450
Opus 5 $0.00225 $0.00225
Sonnet 5 $0.00090 $0.00090
Haiku 4.5 $0.00045 $0.00045

Measured 2d ago against content hash d39072e065cd, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

agent-memory 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.

rules/common/agent-memory.mdc · 75 lines

What it actually says

Must consult existing memory

MUST DO WITH EACH REQUEST

  • First action for each chat request should be to read @agent-memory.md
  • Last action for each chat should be to update @agent-memory.md if needed.

Objective

Ensure Cursor avoids repeating known mistakes by persistently logging corrections and learning. All requests must reference stored knowledge in:

  • docs/agent-memory.md — for known mistakes and their fixes

Rule: Learn from Mistakes

On Every mistake or deprecation warning

  1. Detect incorrect or suboptimal outputs.

  2. Correct the mistake.

  3. Save both mistake and correction in @agent-memory.md using the format:

    ### Mistake: [Short Description]
    **Wrong**:
    [Insert incorrect code or logic]
    
    **Correct**:
    [Insert corrected code or logic]
    

Always add a step to read @agent-memory.md in your thoughts

Rule: Prevent Repetition

On Every Request

  • Always read @agent-memory.md before applying logic.
  • If a related mistake is found:
    • Use the corrected approach instead.
    • Log that a prior fix was applied (optional debug).

Rule: Keep Memory Clean and Updated

  • If a better fix is found for an existing issue:
    • Update that section in @agent-memory.md.
  • Structure the memory files with clear ### headers.
  • Group by topic if needed.
  • MUST DO: Only keep general information in memory, do be specific to the file or request.
  • MUST DO: Avoid storing one-off, file-specific details that are unlikely to be reusable.
  • MUST DO: Information saved in memory should be useful and reusable.

Storage Paths

  • docs/agent-memory.md: Persistent log of mistakes and their fixes

Enforcement

  • Failing to read or update memory files after a mistake is a critical error.
  • All requests must use and comply with stored knowledge.
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 · 75 lines · 450 tokens per session scan A d39072e065cd

Subscribe to this mod's changes

agent-memory is a cursor rule published in the GitHub repository e-gov/cursor-prompts (34 stars, last pushed 4mo ago), licensed MIT. It adds 450 tokens to every session, about $0.0022 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.