# Organize Project Memory Safely

Project context becomes difficult to find when notes are scattered or stale. This skill organizes authorized working memory while preserving registry ownership and privacy boundaries.

## Install

```bash
npx skillstore add aaron-he-zhu/memory-management
```

## Metadata

- Status: approved
- Slug: aaron-he-zhu-memory-management
- Version: 19.0.0
- Author version: 19.0.0
- Skillstore revision: r2
- Version status: valid
- Tree hash: fe0d6717847c97ba4a9972643186b36b835d41768d5da9cfe06bee1e92bd5523
- Author: aaron-he-zhu
- GitHub username: aaron-he-zhu
- License: Apache-2.0
- Repository: https://github.com/aaron-he-zhu/aaron-marketing-skills/tree/main/protocol/memory-management
- Ref: 0715a6e09ea875c8e28cb705ce87cc83045e69c1
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands, filesystem, network
- Quality score: 78
- Quality tier: bronze
- Public page: https://skillstore.pages.dev/skills/aaron-he-zhu-memory-management
- Manifest: https://skillstore.pages.dev/api/skills/aaron-he-zhu-memory-management/manifest

## Capabilities

- Initialize private runtime memory from repository templates after authorization.
- Review HOT, WARM, and COLD notes with source and freshness guidance.
- Consolidate duplicate non-canonical notes while preserving conflicts and provenance.
- Archive stale working notes with path, hash, and supersession metadata.
- Prepare authorized privacy erasure work with dry-run scope reporting.
- Create handoff checkpoints without changing canonical registry facts.

## Use Cases

- Start a new project memory space: Initialize private memory templates and define the first priorities without storing sensitive production data.
- Prepare a cross-session handoff: Refresh a compact checkpoint with pending work, source offsets, and the next approved action.
- Review an erasure request: Plan a scoped privacy purge, identify affected working notes, and report limitations before any authorized deletion.

## Prompt Templates

### Initialize memory

```
Initialize private project memory from the templates. Show the proposed paths and ask for approval before writing.
```

### Review current context

```
Review our HOT and WARM memory. List current priorities, stale notes, and the source record for each conclusion.
```

### Consolidate duplicates

```
Find duplicate non-canonical notes for this campaign. Propose a consolidation plan that preserves conflicts and source references.
```

### Plan a privacy purge

```
Create a dry-run plan to erase the pseudonymous subject ID Subject-A1B2 from authorized working memory. Show scope and request confirmation before changes.
```

## Limitations

- It does not accept, reject, or directly change canonical registry facts.
- Persistent writes and purge operations require explicit authorization.
- It cannot prove complete erasure from Git history, backups, or external systems.
- It depends on project-local templates, registries, and validation scripts.

## Best Practices

- Ask for explicit approval before the first persistent write.
- Store source references and observation dates with every durable note.
- Treat HOT memory as an index and verify important facts in live registry records.

## Anti Patterns

- Do not store credentials, raw contact data, or production exports in templates.
- Do not treat cached notes as proof of current consent or canonical truth.
- Do not delete ambiguous records without a confirmed scope and authority.

## Security Audit

- Audited at: 2026-07-27T11:27:20.123\+00:00
- Summary: Three documented local Python commands are confirmed as external-command actions for initialization, consent lookup, and artifact validation. The remaining 34 detections are false positives from Markdown formatting, fixed documentation links, metadata URLs, and a redirected git error stream. No prompt injection, data exfiltration intent, or malicious behavior was found in the reviewed content.

## Stats

- Views: 0
- Downloads: 6
- Favorites: 0
- Popularity score: 0