Agent Harness Memory Loop — Four Tiers, Feedback Loop, and Load Gates

TL;DR

• External memory is four tiers in practice — short-term, operational, evergreen, and a feedback loop hardened into rules and footers.
• The harness gates when each tier loads so you keep control without token bloat.

Series: Cursor Agent Harness (Part 2 of 3)
Part 1: Lightweight harness without a microservice — you keep model and mode control
Memory foundation: Four tiers of external memory · External Memory series hub · Governance layer

What is the harness memory loop?

The harness memory loop is how a Cursor agent harness gates when each memory tier loads—short-term chat, operational handoffs, evergreen docs, and feedback hardened into rules—so you keep control without pasting 14k tokens on every question.

Who it is for: practice leads and operators who already use Obsidian, handoff files, or session summaries and want those files to load on governed work, skip on quick Q&A, and write back at session end.

What you will learn: the four-tier map (including Layer 4 feedback), which tier lives in git vs vault, and how session summary lines feed the CSV in Part 3.

The problem: memory everywhere means tokens everywhere

External memory works when agents read the same files in the same order every session. It fails when every tier loads on every turn: vault Session Summaries, brain-pack gotchas, MCP Obsidian reads, and subagents each re-embedding 14k tokens for a question that needed one paragraph.

The harness from Part 1 is not only when to spawn subagents. It is when to load which memory tier — while you still own the files, the model choice, and the ship decision.

Why four tiers (not three)

The External Memory Series describes the stack as three layers for clarity. In practice there are four tiers — and the fourth is the one most diagrams label "feedback loop."

Layer 4 is not another folder to grep. It is the Session Continuity System made concrete: session context at the top, v3.1 footer at the bottom, self-improvements with an exact file path, and hooks that regenerate brain-pack.md without you re-explaining gotchas every chat.

brain-pack is a warm embed — mostly Layer 2 facts plus pointers into Layer 3. It is not a substitute for Layer 4. The harness prevents treating brain-pack as "load everything always."

How harness gates map to tiers

You stay in control: the gates are defaults the agent follows. Override with plain language ("skip vault," "read Features/Referrals only," "inline this, no batch").

Rule of thumb: native Read/Grep on vault paths you already have locally. Do not stack MCP + native Read on the same file in one turn.

Git vs Obsidian vs feedback (Layer 4)

Harness policy in git versions with the app. Session handoff in Obsidian stays human-readable. Feedback loop = lessons that graduate into rules (footer Self-improvements must cite path:Lnn or the write did not happen).

Additional detail

Session summary line (Layer 2 → Layer 4 bridge)

Append to Operations/Session Summaries.md:

1- YYYY-MM-DD HH:MM (Project — **title**) — Harness: direct|batch|test|release; tests_before_ship: yes|no; failure: none|wrong-fix|redeploy; — <outcome>

Mirror one row in docs/harness-session-log.csv for weekly math (Part 3). The Summary line is Layer 2 narrative; the CSV is Layer 4 instrumentation you review to decide whether to build more orchestration.

Enable this on another repo

1. Copy context-budget.mdc and HARNESS-POLICY.md from your reference repo.
2. Point sessionStart at your session-prep script; generate brain-pack from memories/repo/.
3. One Obsidian vault root per project in MCP or native Read.
4. Teach agents: Mode A skips bootstrap; ship sessions write one Summary line.

Reference

Quick reference: memory tiers and gates

Common mistakes

FAQ

Is Layer 4 a folder I grep?

No. Layer 4 is feedback hardened: rules, hooks, footer contract, and CI gates—not another directory to paste into every turn.

How is this different from the three-layer External Memory diagram?

The series uses three layers for clarity; in practice there are four tiers—the fourth is the feedback loop (rules + footer + eval). See External Memory hub.

Can I override a memory gate?

Yes. Plain language works: "skip vault," "read Features/Referrals only," "inline this, no batch." Gates are defaults, not locks.

What goes in git vs Obsidian?

Git: HARNESS-POLICY, eval JSON, CSV schema. Obsidian: session narrative, Bridge, open loops. Both: memories/repo/ mirror where you use it.

Do I need MCP for vault reads?

No. Native Read/Grep on local vault paths is often faster. MCP is for writes or when the agent cannot see the disk path.

Reader action

Audit your last five agent sessions. Count how many loaded brain-pack or vault files for questions that needed no code. Map your stack to four tiers — if Layer 4 is only "hope the model remembers," add footer + one rule file before adding another memory product. Part 3 shows how to log and review that discipline weekly. If you're new to Cursor: 50% off your first month (code JP5ARNKSFI2Q). I may earn usage credits; install directly if you prefer.