Memory Landscape Review

Use when the user wants to review auto-memory, promote durable instructions into CLAUDE.md or local memory, and clean up duplicates or conflicts.

Security Scan

The skill's requested access and instructions are consistent with its stated purpose (reviewing and proposing changes to memory layers); it is an instruction-only skill with no installers, credentials, or surprising behaviors.

Purpose & Capability

The name/description match the runtime instructions: the skill is explicitly about auditing repo memory, local memory, and auto-memory and proposing promotions/cleanup. It does not request unrelated credentials, binaries, or config paths.

Instruction Scope

SKILL.md stays within the stated purpose: gather memory entries, classify, detect duplicates/conflicts, propose grouped reports, and only apply changes after explicit approval. It does not instruct the agent to read unrelated system files, exfiltrate data, or call external endpoints.

Install Mechanism

Instruction-only skill with no install spec and no code files to write or execute; therefore no installation risk or external downloads.

Credentials

No environment variables, credentials, or config paths are required. The inputs (repo memory files, auto-memory state, user preference) are directly relevant to the skill's function.

Persistence & Privilege

always:false and no config-modifying install behavior. The skill can be invoked autonomously (platform default), but it does not request permanent presence or elevated privileges and explicitly requires approval before edits.

Assessment

This skill appears coherent and low-risk: it only proposes changes to memory layers and promises no edits without approval. Before installing, confirm (1) your agent runtime only exposes the memory files you expect it to read, (2) you will review and approve any promotions/edits (do not grant blind write permissions), and (3) you keep backups of durable memory files in case you want to revert changes. If you need higher assurance, ask for a transcript of proposed changes before any writes are applied.

Use this skill to review project memory, local memory, and auto-memory together.

Workflow

1. Gather repo memory files and current auto-memory context.
2. Classify each entry as repo-wide, personal, team-wide, or temporary.
3. Detect duplicates, conflicts, and outdated instructions.
4. Present a grouped report before making any changes.
5. Only apply promotions or cleanup after explicit approval.

Guardrails

• Propose first, edit second.
• Do not guess when an instruction might be personal vs shared.
• Keep transient notes out of durable memory files.

Example Requests

• Review my auto-memory and tell me what belongs in durable memory.
• Find duplicate or conflicting instructions across memory layers.

Inputs

• Repo memory files
• Auto-memory state
• User preference for shared vs local memory

Outputs

• Promotion proposals
• Conflict cleanup report
• No-change recommendations

Success Criteria

• Durable memory candidates are identified clearly.
• Duplicates and stale entries are surfaced.
• No edits happen before approval.

Non-Goals

• Silently editing memory files
• Guessing personal vs shared intent when ambiguous

Source Provenance

Derived from src/skills/bundled/remember.ts.