akunzai/agent-skills / skills/agents-memory
agents-memory
VerifiedRecord, recall, or forget notes in the agent's global or project memory, deciding scope (global vs. project) and tier (short-term candidate vs. long-term durable) when recording. Use on `/agents-memory`, `$agents-memory`, "remember this", "what did I note about X", "recall", or "forget X". Autonomous knowledge capture after solving a problem is agents-md's Prevent Recurrence, not this skill.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 3.8 KB |
| references/cross-agent-locations.md | markdown | 1.7 KB |
| references/project-memory-migration.md | markdown | 5.1 KB |
| references/security.md | markdown | 1.0 KB |
| scripts/proj-memory-path.shexecutable | script | 2.9 KB |
| skill.oms.sig | other | 12.7 KB |
SKILL.md
View on GitHub---
name: agents-memory
description: >-
Record, recall, or forget notes in the agent's global or project memory,
deciding scope (global vs. project) and tier (short-term candidate vs.
long-term durable) when recording. Use on `/agents-memory`,
`$agents-memory`, "remember this", "what did I note about X", "recall",
or "forget X". Autonomous knowledge capture after solving a problem is
agents-md's Prevent Recurrence, not this skill.
license: MIT
metadata:
capabilities: shell, writes-outside-repo, edits-agent-instructions
replaces: to-memory
---
# agents-memory — Explicit Memory
Records one thing the user explicitly asked to remember, and recalls or
forgets short-term notes on request.
## Scope and tier
Two independent decisions, each resolved from an explicit argument, semantic
inference from the request, or by asking when genuinely ambiguous:
- **Scope** — `global` (`~/.agents/...`, read by every agent and project) or
`project` (this repo only).
- **Tier** — `long-term` (durable, timeless, must be explicitly requested) or
`short-term` (dated candidate; the default when ambiguous).
## When tier is long-term
Write directly into the durable target — no intermediate staging or batch review:
- Global: `~/.agents/AGENTS.md`
- Project: `<repo>/AGENTS.md`, where `<repo>` is the git root of the current
working directory (the directory itself outside git) — never the directory
this skill was loaded from.
Before writing, hold the note to this bar:
- **Verified** — proven to work in this workspace, not a guess.
- **Reusable** — likely useful again, not one-off task progress.
- **Stable** — a lasting convention, not a temporary hack.
- Strip time-of-day stamps (e.g. `[HH:MM]`) — durable files are timeless.
Placement — a root line or a topic doc behind a pointer — and the size budget
follow the `agents-md` skill.
Always show the exact addition and get explicit confirmation before writing to a
durable file.
## When tier is short-term
Land the note as its own file — one file per note, not appended to a shared log:
```text
~/.agents/memories/YYYY-MM-DD-<slug>.md # global
~/.agents/memories/projects/<project-name>/YYYY-MM-DD-<slug>.md # project
```
- `YYYY-MM-DD` is today's date.
- `<slug>` is a short, agent-chosen topic slug, so the filename alone identifies
the content without opening the file.
- Resolve `<project-name>` with `scripts/proj-memory-path.sh`;
`--ensure` also creates the directory.
- `<project-name>` preserves the root name exactly: main repo for worktrees,
target outside Git. Same-name projects share memory; rename to isolate.
- For project memory, check `~/.agents/memories/.project-name-migration-v1`.
If absent, read [migration](references/project-memory-migration.md) and ask to
copy **all projects**, mapping each `<name>-<hash6>` to its own `<name>`.
Keep verified backups and conflicting notes; remove old directories only
after verification. Mark completion after cleanup. Otherwise skip legacy scans.
- No retention policy or note resolution markers. Delete obsolete notes only
after explicit confirmation.
- Recall: search these directories directly with `grep`/`ls`.
## Never write
Passwords, tokens, credentials, PII, or unverified speculation — see
[references/security.md](references/security.md) for the full list and how to
redact instead of dropping a note entirely.
## Cross-agent global memory setup
Wiring a newly installed coding agent's own native global memory file to
`~/.agents/AGENTS.md` is a one-time setup task, not a memory-capture action — see
[references/cross-agent-locations.md](references/cross-agent-locations.md) for the
per-agent path table. There is no dedicated script for this; wire the file by
hand, in whatever way fits its format, and back up any existing real file
first.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |