Amm save resume

Skill frankenmason/amm-skills/amm-save-resume

Selective curated long-term memory skill bundle for agentmemory MCP. Works with Claude Code, Codex, Cursor, Gemini CLI, Copilot, Antigravity.

Install

npx -y skills add frankenmason/amm-skills --skill amm-save-resume

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Persist a session-end resume point to long-term memory so the next session can pick up without reconstructing state. Trigger at session end, when the user signals wrap-up ("let's continue tomorrow"), or when the context window is near its limit and a safe handoff is needed.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

3.3 KB, 681 tokens by cl100k_base, as published. Nobody here has run it

AMM — Save Resume Point

Purpose

Most agent hosts lose working context between sessions. A resume-point entry captures the current state + pending list + the single next action + files-to-read, so the next session's first turn can restore working context in one memory_recall.

When to Trigger

User signals wrap-up: "that's enough for today", "let's continue tomorrow", "save state".
Agent self-detects session end approaching (context near limit, long idle, or explicit close).
After a major milestone, to bookmark the handoff for the next cycle.

Apply Conditions

At least one pending item exists (otherwise no need to resume — the session is clean).
The "next first action" is concrete (not "continue working").
A short list of files-to-read captures the minimum context restore.

Execute

memory_save({
  content:  "RESUME POINT (#{{SESSION_ID}}, session end):
             [state]   <one-line summary of current state>
             [pending] <bulleted list of unfinished items, with P0/P1/P2>
             [next]    <single concrete first action for next session>
             [read]    <files next session should open, in order>",
  type:     "workflow",
  concepts: "resume_point, {{SESSION_ID}}, <main_topic>",
  files:    "<comma-separated absolute paths in restore-read order>"
})

Warnings

Do NOT save a resume point when the session finished cleanly (no pendings) — it just adds noise on the next session's start recall.
Do NOT rely on this for long-term pendings (>14 days). Use amm-save-pending with an explicit TTL, or a proper project tracker.
The "next action" must be resume-able without the current session's short-term context. If it would require "ask the user what they meant 3 turns ago", revise before saving.

Example

End of a session working on a database migration:

memory_save({
  content:  "RESUME POINT (#delta-7, session end):
             [state]   migration 0042 drafted, review pending, rollback script untested.
             [pending] P0: test rollback on staging.
                       P1: write the up/down migration for the index.
             [next]    run rollback test — psql staging < migrations/0042_down.sql && verify row counts.
             [read]    migrations/0042_up.sql → migrations/0042_down.sql → docs/migration-playbook.md",
  type:     "workflow",
  concepts: "resume_point, delta-7, migration_0042",
  files:    "/workspace/migrations/0042_up.sql,/workspace/migrations/0042_down.sql,/workspace/docs/migration-playbook.md"
})

amm — umbrella (Bootstrap, references, license)
amm-recall — how the next session retrieves this entry
amm-save-pending — per-item tracking with TTL

Change Log

1.0.0 — Initial release.

Gives 0 of the 12 instructions most memory context skills give in 681 tokens

Counted across 674 of the 847 authors here whose files we hold, read 2026-08-06

inform the user when setup is completein 21 of 674, across 6 files
confirm the draft with the user before writingin 21 of 674, across 6 files
update the agent skills block in place if it existsin 21 of 674, across 6 files
present findings to the userin 20 of 674, across 5 files
write the three docs files from seed templatesin 20 of 674, across 5 files
ask the user about each decision one at a timein 19 of 674, across 4 files
edit CLAUDE.md if it existsin 18 of 674, across 3 files
explore current repo statein 18 of 674, across 3 files
do not overwrite user edits to surrounding sectionsin 18 of 674, across 3 files
back up the original file before overwritingin 16 of 674, across 8 files
keep the memory index under 200 linesin 15 of 674
Provide actionable steps and verificationin 13 of 674, across 2 files

Said here and by no other author read

save resume point using memory_save
include a one-line current state summary
list unfinished items with priority tags
list restore-read files in order
set memory type to workflow
set concepts to resume_point plus session id

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.