Load a previously curated memory context for a task
Loads previously curated memory contexts for a specific task or lists all available curated memories.
/plugin marketplace add elafo/centauro/plugin install elafo-centauro-2@elafo/centauroYou are loading previously curated memory contexts that were created by /centauro:curate.
You are a Curated Memory Loader that:
.centauro/contexts/c4-memory/curated/Curated memories are synthesized context files created by /centauro:curate. They are:
Location: .centauro/contexts/c4-memory/curated/memory_curation_[slug]_[timestamp].md
Mode 1: List Mode (when task not provided)
Mode 2: Match Mode (when task provided)
Check if curated directory exists and find all curated memories:
# Check if curated directory exists
if [ ! -d ".centauro/contexts/c4-memory/curated" ]; then
echo "No curated memories found"
exit 0
fi
# Find all curated memories
find .centauro/contexts/c4-memory/curated -name "memory_curation_*.md" | sort -r
If no curated memories found:
ℹ️ No curated memories found
You haven't created any curated contexts yet.
**To create one:**
```bash
/centauro:curate "your task description"
This will analyze your contexts, apply quality filtering, and create a curated memory.
Stop execution if no curated memories exist.
### Step 1: Determine Mode
**Check if task argument provided:**
If task is empty or not provided: → MODE = "list" Else: → MODE = "match"
---
## MODE 1: List All Curated Memories
**When:** Task not provided (`/centauro:load`)
### Step 1: Read Metadata from Each Curated Memory
For each curated memory file, read frontmatter to extract:
- Original task description
- Creation timestamp
- Quality score (if available)
- Source contexts count
- Component types included
**Use Grep or Read tool:**
```bash
# Extract frontmatter from each file
grep -A 20 "^---$" file.md | grep -B 20 "^---$"
Present all curated memories in a table:
# 📚 Available Curated Memories
Found {N} curated contexts in `.centauro/contexts/c4-memory/curated/`
| # | Task | Date | Quality | Sources | Components |
|---|------|------|---------|---------|------------|
| 1 | Add OAuth2 auth | 2025-01-11 | 0.88 (B+) | 12 | c1,c2,c5 |
| 2 | Refactor payment system | 2025-01-10 | 0.92 (A-) | 8 | c1,c2,c4 |
| 3 | Implement caching layer | 2025-01-09 | 0.85 (B) | 10 | c1,c2,c5 |
...
**To load a curated memory:**
```bash
# By number
/centauro:load 1
# By task description
/centauro:load "OAuth2"
/centauro:load "payment system"
To create a new curated memory:
/centauro:curate "your new task"
Stop execution after displaying list.
---
## MODE 2: Match and Load Curated Memory
**When:** Task provided (`/centauro:load "task description"`)
### Step 1: Find Matching Curated Memories
For each curated memory:
1. Read frontmatter to extract original task
2. Calculate relevance score to user's task:
- Exact match: 1.0
- High keyword overlap: 0.8-0.99
- Moderate overlap: 0.5-0.79
- Low overlap: 0.3-0.49
- No overlap: <0.3
**Simple keyword matching:**
### Step 2: Rank Matches
Sort curated memories by relevance score (descending).
**If no matches found (all scores < 0.3):**
```markdown
❌ No matching curated memories found for: "{task}"
**Available curated memories:**
1. Add OAuth2 authentication (2025-01-11)
2. Refactor payment system (2025-01-10)
...
**Options:**
A) Load one of the above by number: `/centauro:load 1`
B) Create new curated memory: `/centauro:curate "{task}"`
C) List all memories: `/centauro:load`
Stop execution.
If 1 strong match (score ≥ 0.7): Automatically load it (skip to Step 4).
If multiple matches (2+ with score ≥ 0.5):
🔍 Found {N} matching curated memories for: "{task}"
**Matches:**
1. **Add OAuth2 authentication** (2025-01-11)
- Relevance: 95% ⭐⭐⭐⭐⭐
- Quality: 0.88 (B+)
- Sources: 12 contexts
2. **Implement JWT authentication** (2025-01-08)
- Relevance: 78% ⭐⭐⭐⭐
- Quality: 0.82 (B-)
- Sources: 8 contexts
**Which to load? (enter number, or 'none' to cancel)**
Wait for user input.
If user selected by number:
If user entered "none" or cancelled:
Cancelled. No context loaded.
Stop execution.
Once selection confirmed, load the curated memory file:
Read the full curated memory:
Read .centauro/contexts/c4-memory/curated/memory_curation_[slug]_[timestamp].md
Display metadata:
# 📄 Loading Curated Memory
**File:** memory_curation_[slug]_[timestamp].md
**Original Task:** [original task from frontmatter]
**Created:** [timestamp]
**Quality:** [score] (Grade [X])
**Sources:** [N] contexts synthesized
**Components:** [c1, c2, c4, c5]
---
Display content: Show the full content of the curated memory (the synthesized information).
After loading, provide clear summary:
# ✅ Curated Memory Loaded
**Loaded:** memory_curation_[slug]_[timestamp].md
**For task:** "{user's task}"
**Matched:** "{original task}" (relevance: {X}%)
## What's Included
This curated memory synthesizes:
- **c1-instructions:** {N} methodologies
- **c2-knowledge:** {M} domain concepts
- **c4-memory:** {P} past decisions
- **c5-state:** {Q} current constraints
## Ready
You now have quality-filtered, task-focused context for: "{task}"
**This context includes:**
- Cross-component patterns
- Dependencies and relationships
- Implementation guidance
- Provenance of source contexts
**Next steps:**
- Start working on your task
- Use `/centauro:reflect` when done to capture learnings
ℹ️ Curated directory doesn't exist yet
You haven't created any curated contexts.
**To create your first curated memory:**
```bash
/centauro:curate "your task description"
This will:
.centauro/contexts/c4-memory/curated/
### Invalid Selection
```markdown
❌ Invalid selection: {input}
Please enter a number from 1 to {N}, or 'none' to cancel.
❌ Error reading curated memory: {filename}
The file may be corrupted or have invalid format.
**Options:**
1. Try loading a different memory
2. Recreate by running: `/centauro:curate "{original task}"`
| Feature | /centauro:load | /centauro:curate |
|---|---|---|
| Purpose | Load existing curated memory | Create new curated memory |
| Process | Simple file read + display | 3-phase pipeline (scan/extract/synthesize) |
| Performance | Fast (~1-2s) | Slower (~30-120s depending on scale) |
| Use when | Reusing previous curation | First time or new task type |
| Quality gate | N/A (already filtered) | Yes (configurable threshold) |
| Output | Displays existing memory | Creates new memory file |
Workflow:
/centauro:curate "task" → Creates curated memory/centauro:load "task" → Reuses curated memory (much faster)/centauro:curate "task" again → Recreates with latest contexts/centauro:load
Output:
# 📚 Available Curated Memories
Found 3 curated contexts
| # | Task | Date | Quality | Sources | Components |
|---|------|------|---------|---------|------------|
| 1 | Add OAuth2 auth | 2025-01-11 | 0.88 (B+) | 12 | c1,c2,c5 |
| 2 | Refactor payment | 2025-01-10 | 0.92 (A-) | 8 | c1,c2,c4 |
| 3 | Caching layer | 2025-01-09 | 0.85 (B) | 10 | c1,c2,c5 |
To load: /centauro:load 1 or /centauro:load "OAuth2"
/centauro:load "Add OAuth2 authentication"
Expected flow:
Output:
# 📄 Loading Curated Memory
File: memory_curation_add-oauth2-auth_2025-01-11-143022.md
Original Task: Add OAuth2 authentication
Relevance: 95% match ⭐⭐⭐⭐⭐
[Full curated memory content...]
✅ Curated Memory Loaded
Ready to work on: "Add OAuth2 authentication"
/centauro:load "authentication"
Expected flow:
Output:
🔍 Found 2 matching curated memories
Matches:
1. Add OAuth2 authentication (2025-01-11)
- Relevance: 75% ⭐⭐⭐⭐
- Quality: 0.88 (B+)
2. Implement JWT tokens (2025-01-08)
- Relevance: 68% ⭐⭐⭐
- Quality: 0.82 (B-)
Which to load? (enter number, or 'none')
/centauro:load 2
Expected flow:
Output:
# 📄 Loading Curated Memory
File: memory_curation_refactor-payment-system_2025-01-10-150433.md
Original Task: Refactor payment system
[Full content...]
✅ Loaded memory #2
/centauro:load "new task"
Output:
ℹ️ No curated memories found
You haven't created any curated contexts yet.
To create one:
/centauro:curate "new task"
Simplicity First: This command should be simple and fast. It's for reusing existing curated memories, not creating new ones.
Smart Matching: Automatically match user's task to existing curated memories. Show options when multiple matches found.
Clear Communication: Always explain which curated memory was loaded and why it's relevant.
No Analysis Needed:
Unlike /centauro:curate, this command does NO context analysis or quality filtering. It just loads pre-curated memories.
Fast Performance: Should complete in 1-2 seconds (just file reads). No agent orchestration needed.
User Control: Allow user to browse all memories or search by task. Support loading by number or task description.
This command is part of a workflow:
First time: /centauro:curate "task"
Reuse: /centauro:load "task"
After work: /centauro:reflect
Refresh: /centauro:curate "task" again
Typical workflow:
Week 1: /centauro:curate "OAuth2" → Creates memory (120s)
Week 2: /centauro:load "OAuth2" → Reuses memory (2s) ✨
Week 3: /centauro:load "OAuth2" → Reuses memory (2s) ✨
Week 4: /centauro:curate "OAuth2" → Refresh with new learnings (120s)
A successful load operation: