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/vectorize-io/hindsight/hindsight-memorygit clone --depth 1 https://github.com/vectorize-io/hindsightWhat 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 | $0.00304 | $0.00304 |
| Opus 5 | $0.00152 | $0.00152 |
| Sonnet 5 | $0.00061 | $0.00061 |
| Haiku 4.5 | $0.00030 | $0.00030 |
Grade A, and why
hindsight-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.
What it actually says
You have access to Hindsight long-term memory through the Hindsight plugin.
Automatic behavior (no action needed):
- Session memory — at the start of each session, the plugin recalls relevant project memories. Recalled memories appear inside a
<hindsight_memories>block in.cursor/rules/hindsight-session.mdc(a file generated by the plugin; safe to delete, will be regenerated). Use those memories to inform your responses. - Auto-retain — after each task, the plugin stores the conversation transcript to Hindsight for future recall. No action needed from you.
On-demand memory tools (via MCP):
- Use the Hindsight MCP tools (
recall,retain,reflect) when the user asks about past decisions, preferences, or project context that isn't covered by the automatic session memory. - Use
recallto search for specific memories beyond what was injected at session start. - Use
retainto explicitly store important decisions or context that should persist. - Use
reflectto reason over accumulated memories when making architectural or design decisions.
Guidelines:
- When memories conflict with current context, prefer current context and note the discrepancy.
- Do not expose raw memory metadata to the user unless asked.
- Prioritize recalled memories that are directly relevant to the current task. Ignore memories that are not applicable.
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 · 22 lines · 304 tokens per session scan A f6498ac79c58
hindsight-memory is a cursor rule published in the GitHub repository vectorize-io/hindsight (21,822 stars, last pushed 2d ago), licensed MIT. It adds 304 tokens to every session, about $0.0015 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.
Other cursor rules, from other repositories
agent-development
Python agents are typically in folders named agent-py/ or agent/ and use the sdk-python/ package.
pi-en
PI Cognitive AI. Trigger: $pi/coding/dev/code/architecture/API/refactor/debug/bug/error/exception/crash/timeout/performance/optimization/test/compile/git/release/verify/review/CR/product/requirements/ops/growth/design/team/support, or deep/2+ failures/looping/stuck/giving-up/retry/nevermind.
pi
你与用户是伙伴🤝战友🔥亲人❤️利益共同体🎯——目标一致：高质量解决问题。百务皆适，融贯古今中西的通才。.
git-author
Git author identity for commits so GitHub attributes contributions correctly.
tools
(none|read|write|network|exec) and risk (low|medium|high|critical). These drive the permission classifier — be honest.
langgraph
for ANY question about LangGraph, use the langgraph-docs-mcp server to help answer -.