Skip to content
Agentmart
Submit a tool

saving-session-learnings

Use when a work session or major task completes — evaluate performance, extract reusable knowledge, and persist it to memory stores and project folders.

Skill by ruca-radio

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/ruca-radio-saving-session-learnings

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

Use when a work session or major task completes — evaluate performance, extract reusable knowledge, and persist it to memory stores and project folders.

Use cases: productivity

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/saving-session-learnings

What's inside

Skill name
saving-session-learnings
Description
Use when a work session or major task completes — evaluate performance, extract reusable knowledge, and persist it to memory stores and project folders.
License
MIT
Instructions
50 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Saving Session Learnings

Close every session by banking what it taught you. Unsaved learnings get relearned at full cost.

## When

- End of a work session.
- After a major milestone: plan executed, incident resolved, feature merged.

Keep it under five minutes. A learning ritual that slows work gets skipped.

## Step 1: Evaluate Performance

Three questions, answered with evidence from the session (ledger, rulings, test runs):

- **What worked?** Name the practice, not the feeling. ("Pre-flight scan caught two type mismatches before dispatch.")
- **What struggled?** Failures, rework rounds, stalls, wrong first guesses — verbatim from the ledger where possible.
- **What changes next time?** Each struggle maps to one concrete change: a rule, a check, a script, a skill edit. No change means it was an observation, not a learning.

## Step 2: Distill Key Knowledge

Separate session trivia from durable knowledge. Durable = useful in a future session without this one's context:

- Project facts: architecture decisions, environment quirks, where credentials live (never the secrets themselves).
- Rulings made and their reasoning (SDD ledger `Ruling:` lines).
- Failure signatures: symptom → root cause → fix.
- Your human partner's preferences and corrections.

## Step 3: Persist

Two stores, two purposes:

1. **Project folder:** `docs/smartpowers/learnings/YYYY-MM-DD-<topic>.md` — the full evaluation: performance answers, rulings, follow-ups. Commit it with the work.
2. **Memory store:** if your harness has persistent memory, save only the durable cross-session facts — one or two sentences each. Project-specific detail stays in the project; memory is for what travels.

Format for future-you: context line, the knowledge, the evidence. No narrative retelling of the session.

## Step 4: Feed the Loop

If a learning indicts a skill — a rule was unclear, a check was missing — record the proposed edit in the learnings file and apply it via `smartpowers:writing-skills` next session. Learnings that never change the system are trivia.

## Rationalization Table

| Excuse | Reality |
|--------|---------|
| "Nothing new happened this session" | Then write one line saying so. The ritual staying cheap is what keeps it alive. |
| "I'll remember it" | You won't — context resets. The file remembers. |
| "Too busy wrapping up" | Five minutes now beats re-deriving it under pressure later. |

From ruca-radio/smartpowers, MIT.

More from ruca-radio/smartpowers

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Aug 22, 2026 (2 months ago)

From GitHub (fetched Oct 24, 2026).

Details

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

More from ruca-radio

Cite this page

saving-session-learnings skill. Agent Mart. https://agentmart.dev/skills/ruca-radio-saving-session-learnings (last updated 2026-10-24)