AgentStack
Browse Sign in
Browse Why AgentStack Sell Docs
Sign in
SKILL verified MIT Self-run

Ucsd Memory

skill-dbalders-ucsd-skills-library-ucsd-memory · by dbalders

Use and maintain an existing local TritonAI memory vault. Trigger on $memory, remember this, forget this, search memory, what do we know about, summarize this into memory, memory status, or requests to add, update, remove, query, or cite Obsidian-style TritonAI memory notes.

No reviews yet
0 installs
14 views
0.0% view→install

Install

$ agentstack add skill-dbalders-ucsd-skills-library-ucsd-memory

✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

View the full security report →

Verified badge

Passed review? Show it. Paste this badge into your README, it links to the public security report.

AgentStack Verified badge Links to your public security report.
[![AgentStack Verified](https://agentstack.voostack.com/badges/verified.svg)](https://agentstack.voostack.com/security/report/skill-dbalders-ucsd-skills-library-ucsd-memory)

Reliability & compatibility

Security review passed
0 installs to date
no reviews yet
2mo ago

Declared compatibility

Claude CodeClaude Desktop

Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.

Preview Execution monitoring

We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.

How agent discovery & health will work →
Are you the author of Ucsd Memory? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

UCSD Memory

Use this skill for day-to-day memory operations after ucsd-memory-create has created the vault.

Locate the Vault

  1. Prefer the configured memory root from .memory/config.json if the user gives a path.
  2. Otherwise check:
  • ~/.agents/ucsd/data/memory
  • %USERPROFILE%\.agents\ucsd\data\memory
  1. If no vault exists, use or recommend ucsd-memory-create instead of inventing a new structure.
  2. Read the vault AGENTS.md before editing memory files.

Operate on Markdown Directly

No wrapper command is required. Inspect and edit the vault's Markdown files directly using the rules below, after reading AGENTS.md and .memory/config.json.

Operations

Remember

Use when the user says to remember, save, capture, add to memory, or keep something for later.

  1. Identify the right scope: person, project, meeting, conversation, preference, workflow, or inbox.
  2. Preserve provenance: source, timestamp, who requested it, and confidence.
  3. Write durable facts to the most specific note. Use inbox.md for uncertain or unsorted memory.
  4. Use YAML frontmatter when creating a new note:
---
title: Example
type: project
created: 2026-06-18
updated: 2026-06-18
sources:
  - kind: conversation
    path: current-thread
confidence: medium
---
  1. Add [[wikilinks]] to related people, projects, meetings, and source notes.
  2. Keep the user-facing response short and mention the file changed.

Search or Answer

Use when the user asks what memory knows, asks about prior work, or asks for a memory-backed answer.

  1. Search filenames, headings, wikilinks, and content.
  2. Prefer specific source-backed notes over broad summaries.
  3. Separate confirmed memory from inference.
  4. Cite the memory note paths or names used.
  5. If memory is stale or ambiguous, say so and offer the narrow next lookup.

Forget or Remove

Use when the user asks to forget, delete, remove, redact, or correct a memory.

  1. Confirm the target if the request could match multiple notes.
  2. Prefer surgical edits or redaction over deleting whole notes.
  3. Preserve a minimal audit line in .memory/logs/ unless the user asks for privacy deletion.
  4. For privacy deletion, remove the sensitive content and do not preserve the sensitive text in logs.
  5. Never remove scheduler config or source files unrelated to the requested memory.

Summarize into Memory

Use when the user asks to save a conversation, summarize a thread, or make the current work durable.

  1. Extract durable facts, decisions, preferences, open loops, and commands.
  2. Put raw or near-raw context under sources/conversations/ only when appropriate.
  3. Update project/person/summary notes with concise source-backed bullets.
  4. Add unresolved items to inbox.md with a Done means: or Follow-up: line when helpful.

Status

Use when the user asks whether memory is working.

Check:

  • vault exists
  • AGENTS.md exists
  • .memory/config.json exists
  • recent .memory/logs/ entries
  • daily and weekly schedule metadata if present
  • enabled sources

Report missing pieces and point setup gaps to ucsd-memory-create.

Guardrails

  • Do not ingest email, calendar, Jira, Drive, or other external systems unless explicitly enabled in the vault config and requested by the user.
  • Do not send email, update Jira, publish files, or make external writes while answering memory requests.
  • Do not store API keys, tokens, passwords, private keys, session cookies, or unrelated secrets.
  • Do not rewrite large areas of the graph for a small memory request.
  • Do not claim memory is current unless you checked the relevant notes or logs in this turn.

Sources & Currency

  • This skill encodes a local operational workflow, not UCSD policy.
  • Pair with ucsd-memory-create for vault initialization and scheduled background jobs.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet, be the first.

Versions

  • v0.1.0 Imported from the upstream source.