codifying-learnings
Use when the user corrects you, repeats an instruction they gave before, or you notice you've made the same mistake or received the same feedback more than…
Skill by Roflochinsky
Chalk Labs
Need help deploying codifying-learnings?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
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/roflochinsky-codifying-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.
About
Reviews
Write a reviewNo 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/codifying-learnings
What's inside
- Skill name
codifying-learnings- Description
- Use when the user corrects you, repeats an instruction they gave before, or you notice you've made the same mistake or received the same feedback more than once.
- License
- MIT
- Instructions
- 41 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Codifying Learnings
## Overview
A correction that lives only in chat history dies at the next compaction — and the user has to repeat it, more annoyed each time. Every correction gets two responses: **fix the instance, then make the rule durable**. Apology without a guardrail is apology theater.
## The loop
1. **Fix the instance** fully (including ripple effects — importers, call sites, tests). Verify, don't just claim.
2. **Sweep**: did you make the same mistake elsewhere in this session's changes? Fix those too.
3. **Pick the durability layer** — the most mechanical one that fits:
| The rule is... | Codify as | Example |
|---|---|---|
| Mechanically checkable | Linter rule / formatter / pre-commit / hook | `import/no-default-export: error` |
| A project convention or fact | Project CLAUDE.md (one line, imperative) | "Named exports only" |
| A personal preference of the user, cross-project | Memory (or user-level CLAUDE.md) | "Explain the why before the detail" |
| A multi-step procedure ("when X, do A then B") | A skill (superpowers:writing-skills) | Release steps, review loop |
| "Every time X happens, run Y" automation | settings.json hook (update-config skill) | Format on save |
Prefer the higher row when several fit: a lint rule beats a CLAUDE.md line beats memory — machines don't forget.
4. **Propose, don't smuggle.** Adding a CLAUDE.md line or memory entry: do it and say so. Changing lint/CI/hook config: propose with the exact change and ask — config edits affect the whole team.
5. **Close the loop out loud**: "Fixed + codified in X, so it won't recur." If you only fixed, say why you didn't codify.
## Trigger discipline
- Trigger on the **first** explicit correction — don't wait for the user to be annoyed twice.
- Also trigger without a correction: if you catch yourself re-learning something you clearly derived in a past session (an env quirk, a build flag, a footgun), codify it now.
- Before writing: check the rule isn't already recorded (CLAUDE.md, memory, lint config). Update the existing entry rather than duplicating; delete entries that turned out wrong.
## Red flags
| Thought | Reality |
|---|---|
| "I'll apologize and fix it" | Fix is half the job. Where does the rule live now? |
| "I'll remember next time" | You won't. Context compacts, sessions end. |
| "Too minor to write down" | Repeated correction = by definition worth one line. |
| "I'll add the lint rule silently" | Team-visible config needs a yes. CLAUDE.md/memory doesn't — just announce it. |From Roflochinsky/workshopskills, MIT.
More from Roflochinsky/workshopskills
- 0.474score
classifying-tasks
Roflochinsky
Use when receiving any new work request, before writing code or choosing an approach — especially multi-item requests, vague one-liners, "quick" asks, or work…
SkillCFreshness: CClaude Code +9 - 0.474score
decomposing-to-issues
Roflochinsky
Use when breaking a spec or plan into tracker issues for execution — especially when different agents or sessions will pick the pieces up independently.
SkillCFreshness: CClaude Code +9 - 0.474score
delivering-heavy-features
Roflochinsky
Use when a task is classified heavy — a new feature with its own business logic, work spanning several layers with a new or changed contract, or anything…
SkillCFreshness: CClaude Code +9 - 0.474score
doctor
Roflochinsky
Use before running heavy composite workflows (delivering-heavy-features etc.), when a required skill or tool is missing, when setting up a new machine or…
SkillCFreshness: CClaude Code +9 - 0.474score
fixing-incidents
Roflochinsky
Use when production is down or degraded, users are reporting errors, or the user demands an urgent fix ("prod is down", "fix it now", "clients are calling") —…
SkillCFreshness: CClaude Code +9 - 0.474score
getting-second-opinion
Roflochinsky
Use when about to approve, merge, or sign off on non-trivial work you authored yourself — especially when feeling confident, when the user says "I trust your…
SkillCFreshness: CClaude Code +9 - 0.474score
onboarding-codebase
Roflochinsky
Use when working in a repository for the first time, or in an unfamiliar area of a large repo — before making any edit, especially when the task "should be…
SkillCFreshness: CClaude Code +9 - 0.474score
upgrading-dependencies
Roflochinsky
Use when asked to update, upgrade, or bump dependencies, fix dependency vulnerabilities, or when CI flags outdated/vulnerable packages — before running any…
SkillCFreshness: CClaude Code +9 - 0.474score
shipping-a-release
Roflochinsky
Use when implementation is complete and the user explicitly asks to wrap up, ship, release, merge, or publish the work.
SkillCFreshness: CClaude Code +9 - 0.471score
heavy-task-pipeline
Roflochinsky
Use for Tier 2 / heavy Workwatch tasks — a new dashboard section with its own business logic, anything spanning API + DB migrations + UI, or any ADR-gate…
SkillCFreshness: CClaude Code +9 - 0.471score
runtime-sql-review
Roflochinsky
Mandatory write→review loop for runtime-/live-only fixes the test suite cannot catch — raw SQL via SQLAlchemy text() / asyncpg, migrations, queries, any…
SkillCFreshness: CClaude Code +9 - 0.445score
grill-me
Roflochinsky
Interview the user relentlessly about a plan or design until reaching shared understanding, resolving each branch of the decision tree.
SkillCFreshness: CClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Jun 16, 2026 (4 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/Roflochinsky/workshopskills
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from Roflochinsky
- 0.474score
classifying-tasks
Roflochinsky
Use when receiving any new work request, before writing code or choosing an approach — especially multi-item requests, vague one-liners, "quick" asks, or work…
SkillCFreshness: CClaude Code +9 - 0.474score
decomposing-to-issues
Roflochinsky
Use when breaking a spec or plan into tracker issues for execution — especially when different agents or sessions will pick the pieces up independently.
SkillCFreshness: CClaude Code +9 - 0.474score
delivering-heavy-features
Roflochinsky
Use when a task is classified heavy — a new feature with its own business logic, work spanning several layers with a new or changed contract, or anything…
SkillCFreshness: CClaude Code +9 - 0.474score
doctor
Roflochinsky
Use before running heavy composite workflows (delivering-heavy-features etc.), when a required skill or tool is missing, when setting up a new machine or…
SkillCFreshness: CClaude Code +9 - 0.474score
fixing-incidents
Roflochinsky
Use when production is down or degraded, users are reporting errors, or the user demands an urgent fix ("prod is down", "fix it now", "clients are calling") —…
SkillCFreshness: CClaude Code +9 - 0.474score
getting-second-opinion
Roflochinsky
Use when about to approve, merge, or sign off on non-trivial work you authored yourself — especially when feeling confident, when the user says "I trust your…
SkillCFreshness: CClaude Code +9
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get codifying-learnings running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
codifying-learnings skill. Agent Mart. https://agentmart.dev/skills/roflochinsky-codifying-learnings (last updated 2026-10-24)