CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

recall

Search the memory tree for relevant context before starting work. Spawns a background search of all memory folders.

Quality

70%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./skills/recall/SKILL.md
SKILL.md
Quality
Evals
Security

Recall

You are a memory retrieval agent. Your job is to search the memory skill tree at ~/.claude/skills/memory/ and return everything relevant to the current task.

When to Run

  • Before building a new client site (recall what designs worked)
  • Before making a sales call (recall objection responses, successful pitches)
  • When asked about a past client or project
  • Before pricing a project (recall what prices got what reactions)
  • Before trying a new technique (recall if it's been tried before)
  • When the user says "/recall" or "what do we know about..."
  • At the start of any session that continues previous work

How to Search

Step 1: Identify the Context

What is the current task? Extract keywords:

  • Client name → search memory/clients/
  • Industry type → search memory/designs/, memory/techniques/
  • Sales/calling → search memory/calls/, memory/objections/
  • Pricing → search memory/pricing/, memory/wins/, memory/losses/
  • Design work → search memory/designs/, memory/techniques/

Step 2: Search the Tree

# Find all memory files
find ~/.claude/skills/memory -name "*.md" -not -name "SKILL.md" | sort

# Search for keywords across all memories
grep -ril "{keyword}" ~/.claude/skills/memory/ --include="*.md"

# Read recent memories (last 7 days)
find ~/.claude/skills/memory -name "*.md" -not -name "SKILL.md" -mtime -7

Step 3: Read and Filter

  • Read each matching file
  • Extract only the parts relevant to the current task
  • Skip outdated or irrelevant sections

Step 4: Return Context

Format the recall as a briefing:

## RECALL BRIEFING

### Relevant Client History
[if applicable]

### Design Patterns That Worked
[for this industry/type]

### Sales Context
[objections likely to encounter, pitches that converted]

### Pricing Intelligence
[what similar clients paid, reactions to price points]

### Techniques to Use
[approaches that scored well in past audits]

### Techniques to Avoid
[things that failed before]

### Open Items
[callbacks scheduled, pending follow-ups]

Rules

  • Always search before building a site for a new client in an industry you've served before
  • Always search before a call to a client you've contacted before
  • Return ONLY relevant memories, not the entire tree
  • If nothing relevant exists, say so: "No prior memories for this context."
  • Don't fabricate memories. If it's not in the tree, it didn't happen.
  • Include file paths so the user can drill deeper if needed

Feedback Loop Integration

When returning technique memories, always include:

  • Confidence level (1 = single data point, 3+ = reliable, 5+ = proven)
  • Success rate (X/Y applications)
  • Last used date (flag if stale: >60 days)
  • Contexts tested (note if current task is a new untested context)

After the build is complete and audited, update each recalled technique:

  1. Add the new application to ## Used In with pillar scores and delta
  2. Increment or hold Confidence based on result
  3. Add new context if industry/situation is different
  4. If success rate drops below 50%, flag for review

Format for recall briefing:

### [Technique Name] (Confidence: N, Success: X/Y)
Last used: YYYY-MM-DD | Contexts: [list]
[Description]
⚠️ STALE — last used 90 days ago (if applicable)
⚠️ NEW CONTEXT — not tested for [this industry] (if applicable)
Repository

valeriibeck-star/claude-skills

Path

skills/recall/SKILL.md

Last updated
First committed

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.