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.
git clone --depth 1 https://github.com/HoangP8/toklessWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/rules/hoangp8/tokless/project-memory)<a href="https://agentmods.dev/rules/hoangp8/tokless/project-memory"><img src="https://agentmods.dev/badge/rules/hoangp8/tokless/project-memory/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/rules/hoangp8/tokless/project-memory"><img src="https://agentmods.dev/badge/rules/hoangp8/tokless/project-memory.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00563 | $0.00563 |
| Opus 5.5 | $0.00225 | $0.00225 |
| Sonnet 5.5 | $0.00113 | $0.00113 |
| Haiku 4.5 | $0.00056 | $0.00056 |
Grade A, and why
project-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 11d 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.
How it starts
The opening of the file, as written. The whole thing — 40 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Project Memory (projectmem)
MCP tools record what broke, what was tried, and what was decided, across sessions, in .projectmem/.
Rules
| Event | Tool | Rule |
|---|---|---|
| Bug found | log_issue |
Trigger: bug, unclear behavior, hypothesis, or investigation topic. |
| Fix tried | record_attempt |
outcome: worked, failed, or partial. One call per attempt. |
| Fix proven | record_fix |
Needs evidence: test passes, error gone, or user confirms. Closes issue. |
| Design choice | add_decision |
Replaces older decision: set supersedes to old event id — the old one retires from the summary, history is never rewritten; two disagreeing decisions corrupt trust. |
| Gotcha found | add_note |
Prefix gotcha:. Shared across projects. Also use for notes and key files future contributors should read first. |
- Log as soon as a trigger fires — before any other work.
- One call per event, decisions and gotchas alike; never batch writer calls.
- Use only the fixed event vocabulary; never invent new event types.
- Log every fix, small ones too. Keep entries short and specific: file, error, test name.
- All writers accept
location: "file:line". - Never log secrets, credentials, customer data, tokens, or large transcripts.
- Edit
PROJECT_MAP.mdandplan.mddirectly. Never log plans as events — a user's idea or plan belongs inplan.md. - Never edit
summary.md(derived) orevents.jsonl(append-only). Stale detail there → record the correction with a writer tool. - Superseding a decision? Get the old event id from
events.jsonl, pass it assupersedes. - Trust recorded memory over recall; don't re-derive answers by re-reading
summary.mdor source files. - Before exiting: memory must reflect what actually happened.
Examples
log_issue(summary: "JWT parse fails on expired token", location: "auth/middleware.go:42")
record_attempt(summary: "Moved expiry check before signature verify", outcome: "failed", issue_id: "0001")
record_fix(summary: "Use leeway comparison at boundary", issue_id: "0001")
add_decision(summary: "Use pgx pool over database/sql", supersedes: "<old event id>")
add_note(summary: "gotcha: local CI skips integration tests without TEST_DB set")
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.
- 11d ago First seen · 40 lines · 563 tokens per session scan A 27213061b202
project-memory is a cursor rule published in the GitHub repository HoangP8/tokless (289 stars, last pushed today), licensed MIT. It adds 563 tokens to every session, about $0.0023 per session on Opus 5.5. 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 2026-10-05.
Other cursor rules, from other repositories
swamr-obsidian-brain
Swamr Obsidian Brain — persistent knowledge base that agents read from and write to across phases. Prevents context loss by maintaining architecture decisions, task outputs, learned patterns, and project knowledge in an Obsidian vault. Every agent MUST read relevant notes before starting and write a summary note when…
decision-logger
Two-layer memory architecture for board meeting decisions. Manages raw transcripts (Layer 1) and approved decisions (Layer 2). Use when logging decisions after a board meeting, reviewing past decisions with /cs:decisions, or checking overdue action items with /cs:review. Invoked automatically by the board-meeting…
cursor-memory
Shokunin memory system: persistent context across sessions.
swamr-state-manager
Swamr State Manager — tracks task progress, manages the execution pipeline, handles retries and escalation. Maintains swamr/state.json as the single source of truth for swarm progress.
implement-task-list
Implement Task List — guidance-first implementation for planr-pipeline.
create-features
Create Features from Epic — break down an epic into features for planr-pipeline.