Memory

Structured memory architecture — hierarchical storage, daily logging, weekly compaction, proactive memory hygiene.

Security Scan

Instructions describe a local file-based memory/indexing system and the actions requested (create, read, archive workspace files) are consistent with that purpose.

Purpose & Capability

The name/description match the runtime instructions: the skill focuses on structured local memory stored in workspace files under memory/*. It does not request unrelated credentials, binaries, or external services.

Instruction Scope

SKILL.md explicitly directs the agent to create, read, append to, archive, and prune files under memory/ (daily/, projects/, people/, decisions/, archive/, and MEMORY.md). All referenced file actions are coherent with a local memory/indexing feature and do not reference unrelated system paths, environment variables, or external endpoints.

Install Mechanism

No install spec or code files are present—this is instruction-only, so nothing is written to disk by an installer. Risk from install-time downloads is absent.

Credentials

The skill requires no environment variables, credentials, or config paths. File access is limited to workspace-relative memory/* paths described in the instructions, which is proportionate to the stated purpose.

Persistence & Privilege

The skill does not request always:true and uses the platform default allowing autonomous invocation. It will create and maintain persistent workspace files (its intended behavior). This persistent file activity is expected for a memory skill but means data written will remain in the workspace.

Assessment

This skill will create and modify files under memory/ and maintain a persistent MEMORY.md index in your workspace. Before installing, ensure you are comfortable with persistent logs and the agent having permission to write to the project directory. Avoid storing highly sensitive secrets in these logs. Test the skill in a throwaway workspace first if you want to inspect its behavior, and consider backing up any important repository data before enabling automatic compaction or pruning.

Memory Architecture

You cannot trust your context window across sessions. Your files are your real memory. Write everything down.

Heartbeat

1. Daily log exists? If no memory/daily/YYYY-MM-DD.md for today → create it
2. Stale index? If MEMORY.md not updated in 3+ days → scan recent daily logs, promote recurring patterns
3. Weekly compaction due? Sunday/Monday → scan last 7 daily logs, promote patterns (3+ occurrences), update project/people files, prune stale entries from MEMORY.md, archive old decisions
4. Orphaned memories? Projects not referenced in 14+ days → flag for archive
5. Storage health? If memory/daily/ has >90 files → archive months older than 60 days into memory/archive/YYYY-MM/
6. If nothing needs attention → HEARTBEAT_OK

Structure

Create and maintain these directories:

• memory/daily/ — one file per day (YYYY-MM-DD.md), append-only session logs
• memory/projects/ — one file per active project (status, goal, decisions, open questions)
• memory/people/ — one file per key person (role, relationship, preferences, history)
• memory/decisions/ — one file per important decision (context, options, rationale, consequences)
• memory/archive/ — compacted old daily logs by month

MEMORY.md — The Index

MEMORY.md in the workspace root is your index. Not a dump. Maximum 200 lines. Every entry links to its detail file.

Sections: Active Projects, Key People, Recent Decisions, Patterns & Lessons, Open Loops.

Promotion rule: Only add to MEMORY.md when something appeared 3+ times in daily logs. Pruning rule: Remove anything not referenced in 30 days.

Session Discipline

Session start: Read MEMORY.md → read today's daily log → read relevant project/people files.

Session end (or /new): Update today's daily log with session summary → create decision/project files if needed → update MEMORY.md if warranted.

Daily Logs

Each memory/daily/YYYY-MM-DD.md has: sessions (timestamp + topic + outcome), tasks completed, tasks created, notes, open loops. Append-only — never edit previous entries. Keep each session entry to 2-4 lines.

Weekly Compaction

Run Sunday/Monday during quiet hours. Scan last 7 daily logs. Extract patterns that appeared 3+ times. Promote to MEMORY.md. Roll up daily notes into project/people files. Prune MEMORY.md entries older than 30 days without reference. Log what was promoted, archived, or pruned.

Rules

• Write everything down. If it matters enough to say, it matters enough to log.
• Index, don't dump. MEMORY.md points to details. It is not the details.
• Daily logs are cheap. Write freely. Compaction handles the noise.
• Project files are authoritative. When daily logs and project files disagree, trust the project file.
• Prune aggressively. Dead entries waste context tokens every session.
• Link everything. Use relative paths so files are navigable.
• Never delete daily logs. Archive them. They're your audit trail.