- Home
- Skills
- Technical Documentation
- Session Learnings to Docs: Stop Your AI Relearning
More screenshots
Works with the AI Agents you already use
Session Learnings to Docs: Stop Your AI Relearning
Stop your AI relearning the same trap. At session end it writes the lessons into CLAUDE.md, AGENTS.md or your repo docs.
$5
Session Learnings to Docs: Stop Your AI Relearning
Example session with this skill installed
We're done for today. Write down what we learned so the next session doesn't have to figure it out again.
- Read your context and instructions
- Compiled the session learnings to
Wrote down what the next session needs
CLAUDE.md: two rules. Run tests withpython3 -m unittest discover, because pytest isn't installed. Never bucket invoices by month on raw UTCcreated_at; convert tosettings.BUSINESS_TZfirst.docs/exports.md: months are in the business timezone, not UTC. It also says why: stored timestamps are UTC, so a late-evening Chicago invoice on the last day of a month falls in the next month in UTC. INV-1042 was the case that exposed it.README.md: the test command saidpytest, which isn't installed and contradicts "No dependencies". It now sayspython3 -m unittest discover.invoicer/settings.py: theBUSINESS_TZcomment said it was only used on printed invoices. It now mentions month assignment in exports. I changed only that comment.
Left out on purpose: the step-by-step of the fix and the regression test. The code, the test and git history already show them.
Check: nothing. I ran python3 -m unittest discover after the edits and it passes.
Not committed. Review with git diff.
Connects securely to your tools. The creator never sees your data.
About this skill
The problem. Every new AI coding session starts from zero. The trap you hit yesterday, the decision you made and why, the test command that actually works: unless someone wrote them down where the next session looks, it rediscovers them the slow way or makes the same mistake again. This skill writes them down at the end of the session, in your repo, so the next session (yours, a teammate's or another agent's) starts from them.
What it does. Say "We're done for today. Write down what we learned so the next session doesn't have to figure it out again." In the example session, that produced two rules in CLAUDE.md (the test command that works, and a timezone rule for monthly invoices), the reason behind the timezone rule in docs/exports.md, the README's wrong pytest command replaced with the one it ran, and a stale code comment corrected. Nothing was committed; the report ended with "Review with git diff".
- Keeps only what a newcomer would get wrong: traps, decisions with the reason given, commands that worked, behavior that changed. Your docs gain rules, not a session diary.
- Puts each item where the next reader looks: the doc section that already covers it, plus a one-line rule in AGENTS.md, CLAUDE.md, Cursor rules or Copilot instructions. The rule loads next session and the detail is one pointer away.
- Fixes docs the session made false, including documented commands it runs and finds broken, so the README stops misleading the next agent.
- Writes commands exactly as it ran them, so the command in your docs is one that worked.
- Leaves everything uncommitted and reports what it wrote, what it left out and what to check, so you review it like any other change.
Why this beats prompting it yourself. The same session, asked the same way.
Without the skill:
- Notes saved only to the agent's own memory, on one machine
- README still told people to run
pytest, which isn't installed - No file in the repo changed
With the skill:
- The test command and the rule in
CLAUDE.md, the reason in the area doc - README fixed after it ran the right command
- A stale code comment corrected too
The notes were accurate both ways. The difference is where they land, so teammates and other agents read them, and the stale docs that get fixed on the way. When a session taught nothing worth keeping, it changes nothing and says so.
Limits. Best run in the same session as the work: in a fresh session it sees only the diff and git log, not the conversation. It edits the docs you have; it does not generate a docs site or API reference. "Never write secrets" is an instruction to your agent, not a scanner, so read the diff before you commit.
What's inside: SKILL.md (the method in five steps, from evidence to report), references/agent-files.md (which instruction file each agent reads, and what belongs in one), examples/worked-example.md (a full session, what was kept or dropped, the edits it produced, and the empty case). Customize it by editing SKILL.md: add your team's doc folder, or name your CHANGELOG format.
$5. Pays for itself the first time a fresh session starts from a rule in your repo instead of hitting the same trap again.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Technical Documentation.
- It writes what the session shows. Run in a fresh session, it sees only the diff and git log, not the conversation, so traps and reasons may be missed.
- "Never write secrets" is an instruction to the agent, not a scanner. Review the diff before committing.
- It edits the docs you have. It does not generate a documentation site or API reference from code.
- The next session benefits only if it reads the files: instruction files load automatically in their agents; other docs are read when the agent looks.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 1 month ago
Permissions
File Scopes
no credentials, no environment reads, no installs. It is told never to write secrets into docs and never to run deploys or migrations; those are instructions to your agent, not locks, so review the diff before you commit.