Personal Preference Learning

Extract preferences, styles, and behavioral patterns from the current conversation and persist them for all future sessions.

Scope: This skill captures how the user works, writes, and wants agents to behave. It does NOT capture code patterns, error fixes, or debugging techniques.

Three-Tier Target Architecture

~/CLAUDE.md

~/.claude/rules/*.md

~/.claude/projects/<project>/memory/

~/CLAUDE.md

~/.claude/rules/

Workflow

Step 1: Scan conversation

Review all user messages for extractable patterns. Use the phrase families below as signals — do not infer preferences from silence or context alone.

Explicit rules (highest priority, save immediately):

• "always…", "never…", "from now on…", "I prefer…"
• "remember that I…", "make sure you…"

Corrections (save if repeated, see Step 2):

• "don't do X", "stop doing Y", "that's not right"
• "actually, it should be…", "I told you before…"
• User rewording or editing Claude's output in place

Positive reinforcement (save when the approach was non-obvious):

• User thanking Claude for a specific behavior
• Accepting an unusual choice without pushback (quieter signal — note why it was non-obvious)

Domain context (save as user-tier facts):

• Role, expertise, institutional affiliation, teaching approach
• Tools, frameworks, or workflows the user commits to long-term

Style edits (save as style rules):

• User modifying prose, formatting, or structure Claude produced
• Terminology corrections ("use X not Y")

Non-signals — skip these:

• One-time instructions ("do X now", "for this file only")
• Hypotheticals ("what if I wanted…", "could you also…")
• Third-party preferences ("my coauthor likes…", "reviewers want…")
• Silence (user didn't push back ≠ user approved)
• Code error resolutions and library workarounds (covered by

  continuous-learning

  )
• Patterns Claude's auto memory already captured this session

Step 2: Apply the repetition threshold

A single off-hand correction is not a durable preference. Before promoting any pattern to later steps:

• Explicit rule ("always…", "never…", "from now on…") → save immediately
• Repeated correction (same pattern corrected 2+ times this session or referenced as "I told you before") → save as tentative, note the repetition
• One-time correction with no repetition and no explicit framing → skip, do not fabricate a rule
• Inferred from silence or hypothetical → skip unconditionally (see

  ## Anti-Patterns

  below)

If a tentative pattern has been corrected 3+ times across sessions (check existing

~/CLAUDE.md

Step 3: Read current state

1. Read

   ~/CLAUDE.md

   in full. Parse the section hierarchy (it evolves — always re-read).
2. List files in

   ~/.claude/rules/

   to check existing rule files.
3. Read the current project's

   ~/.claude/projects/<project>/memory/MEMORY.md

   if it exists.
4. Identify what is already captured across all three tiers to avoid duplication.

Step 4: Detect conflicts with existing preferences

For each extraction, search the already-loaded

~/CLAUDE.md

1. Do NOT silently overwrite the existing rule.
2. In Step 6's preview, show both entries side-by-side under a

   ## Conflicts

   heading and ask which supersedes which.
3. Resolution defaults: more specific scope wins (project > user rule >

   ~/CLAUDE.md

   ); within the same tier, the newer instruction wins only after explicit approval.

Step 5: Classify each extraction

For each pattern that survived Steps 1–4, decide its tier:

~/CLAUDE.md

##

~/.claude/rules/new-topic.md

Categories and typical tiers:

~/CLAUDE.md

## Academic Writing

~/CLAUDE.md

## Email Drafting

~/CLAUDE.md

## Communication

~/CLAUDE.md

~/.claude/rules/workflow.md

~/CLAUDE.md

~/.claude/rules/agent-behavior.md

~/.claude/projects/<project>/memory/

Step 6: Present proposed changes

Show a numbered preview grouped by tier:

## Tier 1: User Instructions (~/CLAUDE.md)

### [1] Add to "## Academic Writing > ### Style" (existing section)
+ - Do not use semicolons to join independent clauses in academic prose

### [2] New section: "## Workflow" (after last existing ## section)
+ - When re-running a pipeline, first check what outputs already exist in cache
+ - Prefer flat directory structures keyed by ID over nested semantic hierarchies

## Tier 2: User Rules (~/.claude/rules/)

### [3] New file: ~/.claude/rules/communication.md
+ # Communication Preferences
+ - Lead with the answer, not the reasoning

## Tier 3: Project Memory

### [4] Add to project memory MEMORY.md
+ - TGL project uses SDT framework (autonomy, competence, relatedness) scored 1-7

## Skipped

- "Use precise terminology" → already in ~/CLAUDE.md ### Word Choice

Approve all? Or specify numbers to modify/reject.

Step 7: Apply approved changes

Tier 1 —

~/CLAUDE.md

• Use the Edit tool for surgical insertions
• Insert new bullets at the end of the target section, before the next heading
• For new sections, use

  ##

  heading level (same as

  ## Email Drafting

  )
• Report the final line count

Tier 2 —

~/.claude/rules/

• Create new rule files as plain markdown with a descriptive

  # Heading

• Use

  paths

  frontmatter only if the rules apply to specific file types
• One topic per file; use descriptive filenames (e.g.,

  workflow.md

  ,

  communication.md

  )

Tier 3 — Project memory:

• Write to

  ~/.claude/projects/<project>/memory/

• Follow the auto memory format:

  MEMORY.md

  as index, topic files for details
• Use the same frontmatter format as existing memory files (name, description, type)

Confirm all changes applied.

Query Mode

The skill can also answer questions about what has already been remembered, without writing anything. Trigger phrases:

##

~/CLAUDE.md

~/.claude/rules/*.md

Query mode is read-only — only "Forget X" writes, and only after explicit confirmation. Never infer intent: if the request is ambiguous, ask before reading.

Anti-Patterns

Common failure modes this skill must avoid. When any of these fire, skip the extraction and do not save anything.

~/CLAUDE.md

Safety Rules

• ALWAYS show proposed changes and get explicit approval before writing any file
• NEVER delete or modify existing preferences — only add or refine
• NEVER rewrite entire sections
• NEVER log preferences inferred from silence, context, or a single off-hand remark (see

  ## Anti-Patterns

  above)
• NEVER save third-party preferences (e.g., "my coauthor likes…") — only the current user's own preferences
• NEVER store credentials, API keys, health information, or PII, even if the user explicitly asks
• If unsure which tier a preference belongs to, ask the user
• If zero preferences are extracted, say so honestly — do not fabricate rules
• If

  ~/CLAUDE.md

  would exceed 200 lines after changes, warn and suggest moving content to

  ~/.claude/rules/

• New sections in

  ~/CLAUDE.md

  must use

  ##

  level to maintain flat hierarchy
• For project memory, respect the existing MEMORY.md structure and index pattern

Out of Scope

This skill does NOT:

• Track numeric confidence scores or usage-decay timers
• Auto-sync to any external agent product config (Cursor, Gemini, OpenClaw, …)
• Modify its own SKILL.md
• Edit

  CHANGELOG.md

  ,

  README.md

  , or git history
• Capture code error resolutions or library workarounds (that is

  continuous-learning

  's job)