memento-momentum

Original：🇺🇸 English

Translated

Enforces a 'Document-then-Execute' workflow. Use when an agent needs to run shell commands, execute tests, build projects, or perform any task that should favor established task runners (Makefile, npm run) and be logged to .cmds-by-agents/ for auditability.

4installs

Sourceryuheechul/agent-skills

Added on2026-03-14

NPX Install

npx skill4agent add ryuheechul/agent-skills memento-momentum

SKILL.md Content

View Translation Comparison →

Memento Momentum

This skill guides agents to prioritize project-specific, established commands and ensures that non-trivial command executions are documented with their Intent in a dedicated directory for auditability.

Core Mandates

1. Prioritize Established Commands

Before executing any command to build, test, lint, or run the project, you SHOULD first research the repository for established task runners or scripts.

Check for:

package.json

Makefile

tasks.json

Justfile

Taskfile.yml

, or scripts in

scripts/

bin/

, or

tools/

Action: If an established command exists (e.g.,
```
npm run test
```
instead of
```
vitest
```
), you SHOULD prefer it.
Nuance: Standard, simple calls to established task runners (e.g.,
```
npm run test
```
,
```
make build
```
) are considered safe and do not strictly require the creation of a documented script unless they are being modified or are part of a larger sequence.

2. Document Non-Trivial Command Executions

To ensure auditability and persistence, AVOID executing bespoke or complex commands directly. Instead, follow this "Document-then-Execute" workflow for non-trivial tasks.

Workflow: Document-then-Execute

Preparation: Ensure the
```
.cmds-by-agents/
```
directory exists (e.g.,
```
mkdir -p .cmds-by-agents
```
).

Creation: Use

write_file

or an equivalent tool to create a shell script in

.cmds-by-agents/

Naming Convention:
```
<purpose>-<YYYYMMDD>-<HHMMSS>.sh
```
.

Content:

bash

#!/usr/bin/env bash
# Intent: [Short description of why you are running this command]
set -x
[Full command here]

Permissions: Make the script executable (
```
chmod +x
```
).
Execution: Run the generated script directly (e.g.,
```
./.cmds-by-agents/script.sh
```
).

Exceptions

Simple research, navigation, and management commands (e.g.,
```
ls
```
,
```
cd
```
,
```
mkdir
```
,
```
grep
```
,
```
git status
```
) are allowed to be run directly.
Standard, direct calls to established project scripts (e.g.,
```
npm start
```
) may be run directly if they don't involve complex flags or environment setup.

3. Security & Secrets

No Secrets: NEVER embed sensitive information (API keys, passwords, tokens) in the command script.
Execution-Time Secrets: Sensitive environment variables must be provided at execution time, either by assuming they are provided by external tooling or by confirming with the user.
No Leakage: If secrets are being handled, DO NOT use
```
set -x
```
in the script to avoid leaking values in the output.

Source Control

Do NOT commit the
```
.cmds-by-agents/
```
directory unless explicitly instructed. It should be ignored in
```
.gitignore
```
.

Example Workflow

Research: You find
```
"test": "vitest"
```
in
```
package.json
```
.
Simple Action: You may run
```
npm run test
```
directly.

Bespoke Action: If you need to run tests with specific filters and a custom reporter:

Create

.cmds-by-agents/test-filtered-20260312-143005.sh