Skip to content
Agentmart
Submit a tool

memento

A lean session framework that gives a coding agent memory across sessions.

Skill by jomavera

IndexedMIT
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/jomavera-memento

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

A lean session framework that gives a coding agent memory across sessions. Opens and closes traceable work sessions, and distils what was learned into a versioned knowledge base under docs/ai/ — session logs, a project map, a learnings ledger and decision records. Use when opening or closing a work session, bootstrapping docs/ai/ in a repository, briefing on where a project stands, auditing the knowledge base, or answering how docs/ai/ is structured and what belongs in it.

Use cases: agent-building

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/memento

What's inside

Skill name
memento
Description
A lean session framework that gives a coding agent memory across sessions. Opens and closes traceable work sessions, and distils what was learned into a versioned knowledge base under docs/ai/ — session logs, a project map, a learnings ledger and decision records. Use when opening or closing a work session, bootstrapping docs/ai/ in a repository, briefing on where a project stands, auditing the knowledge base, or answering how docs/ai/ is structured and what belongs in it.
License
MIT
Metadata
author: jomavera · version: 1.5.0 · homepage: https://github.com/jomavera/memento-session
Instructions
77 lines in SKILL.md · view on GitHub

Files (19)

  • SKILL.md
  • ADR.md · assets/
  • AGENTS-section.md · assets/
  • DESIGN.md · assets/
  • LEARNINGS.md · assets/
  • PROJECT.md · assets/
  • SESSION.md · assets/
  • config.yml · assets/
  • decisions-README.md · assets/
  • sessions-INDEX.md · assets/
  • conventions.md · references/
  • doctor.md · references/
  • profile-analysis.md · references/
  • profile-code.md · references/
  • project-brief.md · references/
  • session-close.md · references/
  • session-discover.md · references/
  • session-init.md · references/
  • session-start.md · references/

SKILL.md (first lines)

# Memento

A work session has an objective, stages with checks, and a record. Memento makes
that record durable: every session leaves a log, and what proves durable is
distilled into a knowledge base the next session reads before touching anything.

Everything lives in the **target repository** under `docs/ai/`, and everything is
committed. `references/conventions.md` is the single source of truth for layout,
identifiers and rules; where a procedure and the conventions disagree, the
conventions win and the procedure is the bug.

## The procedures

Run these, in this order, per unit of work:

| Procedure | When |
|---|---|
| `references/session-discover.md` | Greenfield entry point: no code yet. Aligns on the design, records it in `DESIGN.md`, seeds `PROJECT.md`. Replaces `session-init` on this path. |
| `references/session-init.md` | Once per repository with existing code, to create `docs/ai/`. |
| `references/session-start.md` | Open a session: load the knowledge, agree the objective, plan the stages. |
| `references/session-close.md` | Verify each stage's check, finish the log, distil what is durable. |
| `references/project-brief.md` | Read-only briefing. Writes nothing. Any time. |
| `references/doctor.md` | Audit the knowledge base for stale or broken records. |

Read the procedure you need in full, then follow it. Do not summarise it from
this file — the detail is the point.

Opening and closing a session is a deliberate act. Never run `session-start` or
`session-close` on your own initiative: if substantial work is starting in a
repository that has `docs/ai/`, mention `session-start` once and let the user
decide. `project-brief` and `doctor` are safe to offer freely.

## The reference material

| File | What |
|---|---|
| `references/conventions.md` | Layout, identifiers, promotion tests, correlation, language rules, profiles. Read it before any procedure. |
| `references/profile-code.md` | What to survey and what counts as a check, for software work. |
| `references/profile-analysis.md` | The same, for data pipelines, models and metrics, where a validated check means reconciliation rather than a passing test. |
| `assets/` | The documents the framework creates: `SESSION.md`, `PROJECT.md`, `DESIGN.md`, `LEARNINGS.md`, `ADR.md`, `config.yml`, the index and README seeds, and `AGENTS-section.md`. |

All paths in this bundle are relative to the bundle root — the directory holding
this `SKILL.md`. There are no variables to substitute.

## Harness context

Some values a procedure needs are not part of this bundle because they differ by
harness: the user's argument, a default document language, a conversation id.
Whatever invokes a procedure supplies them as a short **harness context** block,
for example:

```
Harness context:
- Argument: cut the nightly load below 15 min
- Default document language: English
- Conversation id: abc123      (or: this harness exposes none)
```

Where a procedure says "the argument from the harness context" or "the default
from the harness context", that block is what it means. A value that is absent
is simply unset: fall back as the procedure says, and never write a literal
placeholder into a document.

If you are reading this skill directly, with no such block, treat every value as
unset — the argument is whatever the user just asked for, the document language
resolves from `docs/ai/config.yml` or defaults to English, and
`claude_session_ids` stays empty.

## Language

Instructions in this bundle are always English. The **documents the framework
generates** follow a configured language, English by default, resolved per
conventions §9. The machine layer — frontmatter keys, status values,
identifiers, file names, slugs — never gets translated, because the framework
reads it back.

From jomavera/memento, MIT.

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Oct 11, 2026 (19 days ago)
Last release
Oct 8, 2026 (21 days ago)

From GitHub (fetched Oct 24, 2026).

Details

Version
1.5.0
License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from jomavera

Cite this page

memento skill. Agent Mart. https://agentmart.dev/skills/jomavera-memento (last updated 2026-10-24)