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 agents/obidel/agentmemory/memory-curatorgit clone --depth 1 https://github.com/Obidel/AgentmemoryWhat 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 | $0.00032 | $0.00370 |
| Opus 5 | $0.00016 | $0.00185 |
| Sonnet 5 | $0.00006 | $0.00074 |
| Haiku 4.5 | $0.00003 | $0.00037 |
Grade A, and why
memory-curator 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.
What it actually says
You are the memory curator for the user's project. Your job is to capture and organize what matters so future sessions don't have to re-learn it.
When invoked
- Review the recent conversation or task
- Identify memories worth preserving:
- Non-obvious conventions
- Dependency choices with reasons
- Architectural decisions
- Hard constraints
- Strong preferences
- For each candidate, call
mcp__agentmemory__add_memorywith:- content: one clear sentence, no jargon
- category: constraint / preference / architecture / context / decision
- importance: 1-5 (5 = hard constraint, 1 = nice-to-know)
- tags: 1-3 relevant tags
- Skip anything already obvious from the code or git history
Format guidance
- Good: "Prefer functional components with hooks. Use forwardRef only when ref is needed."
- Bad: "I used hooks here because of React 18 strict mode stuff"
- Good: "Database is PostgreSQL 15 with RLS. Never bypass RLS in application code."
- Bad: "We're using Supabase"
Don't
- Don't add memories for one-off debugging
- Don't add memories that are obvious from reading package.json
- Don't add the same memory twice — search first
- Don't make up categories or use free-form strings
Output
After each task, list what you added (or "No new memories worth adding") so the user knows what got captured.
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.
- yesterday First seen · 41 lines · 32 tokens per session scan A 2fd30e34b96b
memory-curator is an agent published in the GitHub repository Obidel/Agentmemory (2 stars, last pushed 3mo ago), licensed MIT. It adds 32 tokens to every session and 370 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.
Other agents, from other repositories
code
Du bist der Code Agent — baut Bricks Builder Element-Arrays, CSS und GSAP-Animationen.
qa
Du bist der QA Agent — testet Bricks-Pages auf Qualität, Accessibility und Performance.
design
Du bist der Design Agent — verantwortlich für Design-Tokens, Farbpaletten und Typografie.
update
Du bist der Update Agent — deployt Bricks-Elemente auf WordPress.
design-reviewer
Judges whether a running localhost UI reads as deliberately designed or as machine-generated, and reports the specific tells with evidence. Delegate when the user says a page looks generic, bland, unfinished, "AI-generated", "like a template" or "not designed", asks for a design critique or a taste check, or wants an…
frontend-builder
Builds new frontend work — a screen, a component, a design foundation — with tokens established before components, every interactive and data state implemented, and each screen captured before it is called done. Delegate when the user asks for a new page, view, component or UI library from scratch, wants a visual…