Skip to content
Skills Library
AI & agents · Automation

capture-learning

/capture-learning

Captures a validated learning and writes it to the correct learnings.md. Invoke when: a bug is resolved non-obviously, a pattern is discovered, the user corrects a mistake, or a solution took 2+ attempts. Do NOT invoke for typos, obvious errors, or one-time config changes.

Repository
Synapse ↗
Licence
AGPL-3.0
Last repository change
15.09.2026

Assessment and security

100/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content35 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
2
Scripts
0
HTML comments
2
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 15.09.2026, without references to retired models or APIs.

Licence

AGPL-3.0, copyleft: przeróbki udostępniane innym muszą mieć tę samą licencję.

What to check before installing
  • SKILL.md zawiera komentarze HTML (2). Nie widać ich w podglądzie Markdown, a model je czyta.

What it does and when to use it

Captures a validated learning and writes it to the correct learnings.md. Invoke when: a bug is resolved non-obviously, a pattern is discovered, the user corrects a mistake, or a solution took 2+ attempts. Do NOT invoke for typos, obvious errors, or one-time config changes.

Arguments

/capture-learning descrivi cosa hai imparato e in quale contesto
Original author description

Captures a validated learning and writes it to the correct learnings.md. Invoke when: a bug is resolved non-obviously, a pattern is discovered, the user corrects a mistake, or a solution took 2+ attempts. Do NOT invoke for typos, obvious errors, or one-time config changes.

How it works

  1. When to invoke
  2. Protocol
  3. Multi-agent safety
Original SKILL.md headings · 124 lines, 491 words
SKILL.md preview
---
name: capture-learning
description: >
  Captures a validated learning and writes it to the correct learnings.md.
  Invoke when: a bug is resolved non-obviously, a pattern is discovered,
  the user corrects a mistake, or a solution took 2+ attempts.
  Do NOT invoke for typos, obvious errors, or one-time config changes.
version: 1.0.0
user-invocable: true
argument-hint: "descrivi cosa hai imparato e in quale contesto"
---

# Capture Learning

## When to invoke

- Bug resolved with non-obvious root cause
- Pattern discovered that applies beyond this one case
- User corrected your approach
- Solution took 2+ attempts to get right
- Framework or library quirk discovered

Do NOT invoke for: typos, missing imports, obvious variable errors, one-off config.

## Protocol

### Step 1 — Choose target skill

| Domain | Target |
|--------|--------|
| Debugging, error fixing | `systematic-debugging/learnings.md` |
| Next.js, React, App Router | `next-best-practices/learnings.md` |
| CodeGraph usage | `codegraph-context/learnings.md` |
| UI, components, styling | `frontend-design/learnings.md` |
| Session/workflow patterns | `verification-before-completion/learnings.md` |
| Vercel, React performance | `vercel-react-best-practices/learnings.md` |

If uncertain: use the skill most closely related to the DOMAIN, not the activity.

### Step 2 — Draft the learning

Fill ALL required fields:

```yaml
## Learning L-{skill}-{NNN}
id: "L-{skill}-{NNN}"
date: "{today YYYY-MM-DD}"
type: "{bug-fix|pattern|anti-pattern|preference|negative}"
status: "active"
project: "{project-name or global}"
scope: "{project-specific|global}"
tags: [{tag1}, {tag2}]
confidence: 1
context: "In [technology/version], when [specific situation]..."
problem: "{what went wrong or was unclear}"
solution: "{what worked — actionable, specific}"
reason: "{WHY this works — not just what}"
validated_by: ["{session-date}"]
created_in: "{session-date}"
supersedes: null
superseded_by: null
reinforces: []
contradicts: []
```

### Step 3 — Validate (no exceptions)

Before writing, verify ALL of these:

1. `context`, `problem`, `solution`, `reason` are filled and non-empty
2. `tags` has between 2 and 5 items
3. `problem` and `solution` contain NO absolute file paths
4. `solution` is at least 10 words and actionable
5. `reason` explains WHY (not just restates the solution)
6. If `type: negative` then `confidence` must be 1

Fix any failure before writing. Do not skip any check.

### Step 4 — Contradiction check

Search existing learnings for entries sharing 2+ tags with the new one:

```bash
grep -r "tags:" "/Users/dan/Desktop/progetti-web/Fullstack session/.agents/skills/*/learnings.md" \
  | grep "{tag1}\|{tag2}"
```

If a conflict is found, surface it to the user:

```
CONFLICT DETECTED
New:      "{summary}" [tags: ...]
Existing: L-{id} "{summary}" — confidence: N

A) New supersedes old → set supersedes/superseded_by fields
B) Both valid with different scope → adjust scope
C) Cancel new learning
```

Wait for the user's decision. Never auto-resolve.

### Step 5 — Write

Calculate the next ID: count existing `## Learning` lines in the target file + 1, zero-padded to 3 digits.

Append the learning between `<!-- LEARNINGS START -->` and `<!-- LEARNINGS END -->`.

### Step 6 — Re-index skills with embeddings

```bash
node "/Users/dan/Desktop/progetti-web/Fullstack session/packages/codegraph/dist/cli.js" analyze "/Users/dan/Desktop/progetti-web/Fullstack session/.agents/skills" --skip-git 2>&1 | tail -3
```

This updates the structural index. Semantic search works at query-time via the embedding model (loaded automatically on first query).

### Step 7 — Confirm

```
✅ Learning captured: L-{skill}-{NNN}
   Type: {type} | Confidence: 1 | Scope: {scope}
   Tags: {tags}
   File: {skill}/learnings.md
```

## Multi-agent safety

If running as a subagent, do NOT write directly to `learnings.md`.
Write to a temp file instead:

```
/Users/dan/Desktop/progetti-web/Fullstack session/.agents/skills/_pending/learning-{date}-{random}.yml
```

The parent agent or `post-session-review` will merge these at session end.
Files in this package (2)
  • SKILL.md 4 KB
  • learnings.md 1 KB

Versions

RepositoryScoreLicence
This page · PIXARTSeu/Synapse100AGPL-3.0
arozumenko/sdlc-skills100MIT
PIXARTSeu/Synapse100AGPL-3.0

Database snapshot: 13.11.2026 · package version bbc4573ba1f6 (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 100learning-to-rule-or-skill
    Use when a repeated learning, mistake, or successful pattern should be turned into a new rule or skill. Also use after completing a task to capture learnings from the work.
  • 100learnings
    Capture or recall evidence-backed execution learnings. Use when observed evidence changes a reusable decision, reveals a durable footgun, or corrects a prior learning; also for lessons learned, learning recall, or memory admission and supersession. Delivery alone is not a trigger.
  • 90recoup-platform-capture-lesson
    Capture a solved problem or reusable lesson so the next run is cheaper — the box's compounding memory. Use when the user says "that worked", "remember this", "capture this learning", "save this so we don't redo it", "write this down for next time", or after a non-trivial…
  • 100self-learning
    Continuous self-improvement through systematic logging, pattern detection, and behavioral updates. Use when: the owner corrects you, a task fails, you discover a better approach, or you notice a recurring pattern. Store raw learnings in .learnings/, update the specific skill or…
  • 100capture-learnings
    Capture and apply accumulated knowledge via structured memory. Use when the user gives feedback, shares preferences, corrects a mistake, or when you discover something worth remembering for future conversations.
  • 100learning-capture
    Capture a durable learning by editing context directly. Triggers when the user corrects Claude in a generalizable way ("use X instead of Y", "always X", "never Y", "I prefer X", "from now on X"), states a reusable preference, or asks to "remember this" / "save this". Also…

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.