agentsclimarketplace

Amm save directive

Skill frankenmason/amm-skills/amm-save-directive

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-directive

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 user's explicit directive to long-term memory via agentmemory MCP, so the agent can recall it 4-5 turns later without relying on short-term context. Trigger on phrases like "remember this", "from now on", "always do X", "never do Y", or when the user asks the agent to save something explicitly.

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

3.2 KB, 662 tokens by cl100k_base, as published. Nobody here has run it

AMM — Save Directive

Saves a user-issued directive as a preference-type memory entry.

Purpose

Short-term context can decay within 5–10 turns. A user's "always use X" rule should survive that drift. This skill captures the directive (with topic tag) so future turns can recall it via memory_recall.

When to Trigger

  • User message contains "remember", "save", "record", "from now on", "always", "never" as a rule, not narrative.
  • User explicitly asks "save this to memory".
  • Agent self-detects a directive that will matter 4+ turns later AND is not already derivable from git / codebase.

Apply Conditions

  • The directive is concrete and actionable (not a vague preference).
  • ≥14-day relevance window (ephemeral rules belong in conversation, not memory).
  • Not already saved in the last 10 turns (check with a quick memory_recall first when practical).

Execute

memory_save({
  content:  "User directive (#{{SESSION_ID}}): <verbatim or close paraphrase>.
             Action taken: <what the agent did in response this turn>.
             Files involved: <comma-separated absolute paths or 'none'>",
  type:     "preference",
  concepts: "user_directive, <topic_tag>, {{SESSION_ID}}",
  files:    "<comma-separated absolute paths or empty string>"
})

Optional round-trip verification

Some agentmemory versions lag the search index. If verification matters this turn, prefer a direct file check over memory_recall:

grep -F "<unique phrase from content>" ~/.agentmemory/standalone.json

(Use memory_recall instead if your build's recall is reliably synchronous with saves.)

Warnings

  • Do NOT include secrets, tokens, credentials, or PII in content or concepts. The storage file is plaintext.
  • Do NOT save a directive that will be contradicted by the next directive. Favor higher-level, longer-lived rules.
  • Do NOT duplicate within a session. If the same directive is issued twice in 10 turns, skip the second save.

Example

User says: "from now on, always use uv for Python dependency management on this project — pip has been unreliable."

memory_save({
  content:  "User directive (#alpha-1): always use uv for Python deps on {{PROJECT}}.
             Rationale: pip unreliable on this project.
             Action: replaced pip commands with uv in next script.",
  type:     "preference",
  concepts: "user_directive, python, dependencies, uv, alpha-1",
  files:    "/workspace/scripts/install.sh"
})

Related

Change Log

  • 1.0.0 — Initial release.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.