learn

A command for recording and managing corrections and preferences for coding skills. It can capture new guidance, display stored guidance, remove entries, reset a skill, or combine repeated patterns.

In plain words
What is it for?
Use it after giving feedback about how a skill should behave, to inspect learned preferences, or to clean up and consolidate stored guidance.
Why use it?
It keeps useful feedback from being lost and makes it easier to maintain consistent working preferences across skills.

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/drag88/claude-dev-framework/learn
Clone the repo
git clone --depth 1 https://github.com/drag88/claude-dev-framework
Per session 36 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,328 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 $0.00036 $0.01328
Opus 5 $0.00018 $0.00664
Sonnet 5 $0.00007 $0.00266
Haiku 4.5 $0.00004 $0.00133

Measured yesterday against content hash 2fbef81e243b, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

learn 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 yesterday.

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.

commands/learn.md · 182 lines

How it starts

The opening of the file, as written. The whole thing — 182 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/cdf:learn

Universal skill learning command. One command for capture, viewing, removal, and consolidation.

Usage

/cdf:learn "preference or correction"              # Capture (infers skill)
/cdf:learn skill-name "preference or correction"    # Capture (explicit skill)
/cdf:learn status                                   # Overview across all skills
/cdf:learn show skill-name                          # Show entries for a skill
/cdf:learn remove skill-name number                 # Remove specific entry
/cdf:learn reset skill-name                         # Clear all for a skill
/cdf:learn consolidate skill-name                   # Deduplicate and merge

Subcommands

Capture (default)

/cdf:learn "preference" or /cdf:learn skill-name "preference"

  1. Determine target skill:

    • Explicit name provided → use it (validate exists in skills/)
    • No name → infer from conversation context (which skill was most recently active)
    • Inference fails → ask: "Which skill?" and list recently active ones
    • No skill applies → log to Claude's auto-memory instead (this command is for skill-specific learning)
  2. Classify into section:

    • Do: Positive instructions ("always X", "prefer X", "use X")
    • Don't: Negative instructions ("never X", "avoid X", "stop doing X")
    • Style: Tone, format, voice preferences ("be more concise", "use parentheses not dashes")
  3. Write to skills/{skill-name}/learned.md:

    • Create from template if it does not exist
    • Append dated entry: - {description} (YYYY-MM-DD)
    • Silent. Do NOT announce that feedback was logged.
  4. Check cap: If 20+ entries, notify: "Run /cdf:learn consolidate {skill-name} to merge."

Status

/cdf:learn status

Scan skills/*/learned.md across all skill directories. Show:

Learned preferences across skills:

  frontend-design:   9 entries (last updated 2026-03-28)
  coding-standards:  2 entries (last updated 2026-03-25)
  frontend-patterns: (no learned.md)

Total: 11 entries across 2 skills

Read the full file on GitHub · 182 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. yesterday First seen · 182 lines · 36 tokens per session scan A 2fbef81e243b

Subscribe to this mod's changes

learn is a command published in the GitHub repository drag88/claude-dev-framework (2 stars, last pushed 1mo ago), licensed MIT. It adds 36 tokens to every session and 1,328 once invoked, about $0.0002 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.