agentleFS
hiteshbandhuhiteshbandhu/skills-i-use/skills/ai-engineer-talks/operate-agent-memory/SKILL.md

operate-agent-memory

Skill1 starsChanged 5 months agoMITView on GitHub

Runs checklists and workflows for agent context engineering — separating context from memory, smart truncation, sub-agent offload, and long-session evals. Use when the user hits context limits on observability or trace-heavy agents, designs multi-turn memory, or says "context management", "escape context window", "agent forgets follow-ups".

The file

Files
SKILL.md

Operate agent memory

Action playbook from Sally-Ann Delucia (Arize Alex). Do not summarize the talk — pick a workflow and execute it.

Supporting files:

  • workflows.md — workflows A–E
  • source-index.md — talk → learnings

Optional deliverables: {SKILL_OUTPUT_DIR}/operate-agent-memory/


Step 0 — Pick workflow

What is the user trying to do?
├─ Diagnose context death spiral (retry adds data)     → A
├─ Design in-window vs memory split                    → B
├─ Implement truncation + retrievable store              → C
├─ Add long-session regression evals                     → D
└─ Split main vs sub-agent context budgets               → E

Stop summarizing once a workflow is identified — run its checklist.


Install

cp -r skills/operate-agent-memory ~/.claude/skills/
cp -r skills/operate-agent-memory ~/.cursor/skills/
cp -r skills/operate-agent-memory ~/.codex/skills/

Source: ingest-into-skills playlists/memory-ai-engineer/.


Cross-cutting rules

RuleSource
Context = what the model sees; memory = what survives outside[src-001 @ 7:48]
Agents fail on context, not prompts[src-001 @ 14:14]
Naive head truncation breaks follow-ups[src-001 @ 5:18]
Uncontrolled summarization is unreliable[src-001 @ 6:15]
Long sessions fail late — eval turn N+1[src-001 @ 8:44]

Output to user

  1. Name the workflow (A–E) and deliverable
  2. Save artifacts under ./skill-outputs/operate-agent-memory/ when useful
  3. Do not auto-commit

Invocation examples

@operate-agent-memory our agent loses thread on turn 12
design sub-agents for heavy trace search
long-session eval harness for support copilot

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.