learn-curate

A command that organizes a collection of saved learnings by assigning categories, finding similar entries, merging duplicates, archiving old entries, and updating timestamps. It also has a dry-run mode that previews changes without saving them.

In plain words
What is it for?
Use it to curate accumulated learnings, review duplicate or similar entries, archive learnings older than 50 sessions, or test changes safely with a dry run.
Why use it?
It reduces clutter and repeated notes as the collection grows. The preview mode lets you inspect proposed changes before altering the saved file.

Command

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add commands/ankushdixit/claude-plugins/learn-curate
Clone the repo
git clone --depth 1 https://github.com/ankushdixit/claude-plugins
Per session 5 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 523 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00005 $0.00523
Opus 5 $0.00003 $0.00262
Sonnet 5 $0.00001 $0.00105
Haiku 4.5 $0.00001 $0.00052

Measured 2d ago against content hash 940af23bf8ca, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

learn-curate scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

solokit/commands/learn-curate.md · 97 lines

How it starts

The opening of the file, as written. The whole thing — 97 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Curate Learnings

Run automatic categorization, similarity detection, and merging of learnings.

What Curation Does

The curation process:

  1. Categorizes uncategorized learnings using AI-powered keyword analysis
  2. Detects duplicates using Jaccard and containment similarity algorithms
  3. Merges similar learnings to reduce redundancy
  4. Archives old learnings (learnings older than 50 sessions)
  5. Updates metadata (last_curated timestamp)

Usage

Normal Curation (Save Changes)

sk learn-curate

This will:

  • Process all learnings
  • Save changes to learnings.json
  • Display summary of actions taken

Dry-Run Mode (Preview Only)

sk learn-curate --dry-run

This will:

  • Show what changes would be made
  • NOT save any changes
  • Useful for previewing curation results

When to Run Curation

Manual curation is useful when:

  • You've captured many learnings and want to organize them
  • You want to check for duplicate learnings
  • You want to preview what auto-curation would do
  • You're testing the curation process

Note: Curation also runs automatically every N sessions (configurable in .session/config.json).

Output Format

Display the curation summary showing:

  • Initial learning count
  • Number of learnings categorized
  • Number of duplicates merged
  • Number of learnings archived
  • Final learning count

Example output:

=== Learning Curation ===

Initial learnings: 45

✓ Categorized 8 learnings
✓ Merged 3 duplicate learnings
✓ Archived 2 old learnings

Final learnings: 42

✓ Learnings saved

Understanding the Process

Categorization: Uses keyword analysis to assign learnings to one of 6 categories:

  • architecture_patterns, gotchas, best_practices, technical_debt, performance_insights, security

Similarity Detection: Uses two algorithms:

  • Jaccard similarity: Measures word overlap (threshold: 0.6)
  • Containment similarity: Detects if one learning contains another (threshold: 0.8)

Merging: Combines tags and tracks merge history when duplicates are found.

Read the full file on GitHub · 97 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 97 lines · 5 tokens per session scan A 940af23bf8ca

Subscribe to this mod's changes

learn-curate is a command published in the GitHub repository ankushdixit/claude-plugins (3 stars, last pushed 7mo ago), licensed MIT. It adds 5 tokens to every session and 523 once invoked, about $0.0000 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.