Test Memory

1---
2name: test-memory
3description: Comprehensive interactive testing of all Memory MCP features
4allowed-tools: "mcp__memory__*"
5---
6 
7# Memory MCP Live Testing Skill
8 
9Run `/test-memory` to start a guided testing session. Walk through each phase, execute tests, and track results.
10 
11## Prerequisites
12 
13Install memory-mcp from the marketplace:
14 
15```bash
16# Add the marketplace
17/plugin marketplace add michael-denyer/memory-mcp
18 
19# Install the plugin
20/plugin install memory-mcp@michael-denyer/memory-mcp
21```
22 
23Or install directly:
24```bash
25/plugin install github:michael-denyer/memory-mcp
26```
27 
28## Phases
29 
30---
31 
32### Phase 1: Core Memory Operations
33 
34**1.1 Remember** - Store test memories:
35```
36remember("Test: FastAPI with async endpoints", memory_type="project", tags=["test", "tech-stack"])
37remember("Test: Always use uv run pytest -v", memory_type="pattern", tags=["test", "commands"])
38remember("Test: API rate limit 100 req/min", memory_type="reference", tags=["test", "api"])
39```
40Record the returned IDs for later tests.
41 
42**1.2 Recall** - Semantic search (use different phrasings):
43- "what framework for backend" → should find FastAPI
44- "how to run tests" → should find pytest
45- "request throttling" → should find rate limit
46 
47Verify confidence levels returned.
48 
49**1.3 Recall Modes** - Compare precision vs exploratory:
50- `recall("testing", mode="precision")` → fewer, higher-confidence
51- `recall("testing", mode="exploratory")` → more results
52 
53**1.4 Recall by Tag**:
54- `recall_by_tag("test")` → should return all test memories
55 
56**1.5 Forget** - Delete one test memory and verify it's gone.
57 
58---
59 
60### Phase 2: Hot Cache Mechanics
61 
62**2.1 Manual Promotion**:
63- `promote(memory_id)` one of the test memories
64- `hot_cache_status()` → verify it appears
65 
66**2.2 Manual Demotion**:
67- `demote(memory_id)`
68- Verify removed from hot cache but still recallable
69 
70**2.3 Pin/Unpin**:
71- `pin(memory_id)` → pinned_count increases
72- `unpin(memory_id)` → pinned_count decreases
73 
74**2.4 Auto-Promotion** (optional - requires multiple recalls):
75- Create a memory and recall it 3+ times
76- Check if auto-promoted
77 
78---
79 
80### Phase 3: Knowledge Graph
81 
82**3.1 Create Linked Memories**:
83```
84remember("Test: Database uses PostgreSQL") → ID: A
85remember("Test: pgvector for embeddings") → ID: B
86remember("Test: Vector search needs pgvector") → ID: C
87link_memories(A, B, "relates_to")
88link_memories(B, C, "depends_on")
89```
90 
91**3.2 Traverse Graph**:
92- `get_related_memories(B)` → should show A and C
93- `get_related_memories(A, direction="outgoing")` → should show B
94 
95**3.3 Multi-Hop Recall**:
96- `recall("PostgreSQL", expand_relations=true)` → should include related memories
97 
98**3.4 Unlink**:
99- `unlink_memories(A, B)`
100- Verify relationship removed
101 
102---
103 
104### Phase 4: Trust Management
105 
106**4.1 Validate**:
107- `validate_memory(id, reason="used_correctly")`
108- Check trust_score increased
109 
110**4.2 Invalidate**:
111- `invalidate_memory(id, reason="outdated", note="Testing invalidation")`
112- Check trust_score decreased
113 
114**4.3 Trust History**:
115- `get_trust_history(memory_id)` → shows all changes
116 
117---
118 
119### Phase 5: Contradiction Detection
120 
121**5.1 Create Conflicting Memories**:
122```
123remember("Test: Timeout is 30 seconds") → ID: X
124remember("Test: Timeout is 60 seconds") → ID: Y
125```
126 
127**5.2 Find & Mark**:
128- `find_contradictions(X)` → should suggest Y
129- `mark_contradiction(X, Y)`
130- `get_contradictions()` → pair listed
131 
132**5.3 Resolve**:
133- `resolve_contradiction(X, Y, keep_id=X, resolution="supersedes")`
134- Verify X supersedes Y, Y's trust reduced
135 
136---
137 
138### Phase 6: Sessions & Episodic Memory
139 
140**6.1 Check Sessions**:
141- `get_sessions(limit=5)`
142 
143**6.2 Episodic Memories**:
144```
145remember("Test: Debugging auth today", memory_type="episodic")
146remember("Test: Found token bug", memory_type="episodic")
147```
148 
149**6.3 Session Topic**:
150- `set_session_topic(session_id, "Testing session")`
151 
152**6.4 Summarize Session**:
153- `summarize_session(session_id)` → structured summary with:
154  - Decisions (choices made and rationale)
155  - Insights (lessons, antipatterns, landmines, constraints)
156  - Action Items (todos, bugs, tasks)
157  - Context (background, conventions, preferences, architecture)
158- Use before `end_session()` to review what will be promoted
159 
160**6.5 End Session** (optional - ends current session):
161- `end_session(session_id, promote_top=true)`
162 
163---
164 
165### Phase 7: Pattern Mining
166 
167**7.1 Log Output**:
168```
169log_output("import pandas as pd")
170log_output("import numpy as np")
171log_output("uv run pytest -v")
172```
173 
174**7.2 Run Mining**:
175- `run_mining(hours=1)`
176- `mining_status()`
177 
178**7.3 Review Candidates**:
179- `review_candidates()`
180 
181**7.4 Approve/Reject** (if candidates exist):
182- `approve_candidate(id)` or `reject_candidate(id)`
183 
184---
185 
186### Phase 8: Seeding & Bootstrap
187 
188**8.1 Seed from Text**:
189```
190seed_from_text("- Item one\n- Item two\n- Item three", memory_type="project")
191```
192Verify 3 memories created.
193 
194**8.2 Seed from File** (creates temp file):
195- `seed_from_file(file_path, memory_type="reference")`
196 
197**8.3 Bootstrap Project**:
198- `bootstrap_project(root_path=".", promote_to_hot=false)`
199- Verify it finds CLAUDE.md, README.md, etc.
200 
201---
202 
203### Phase 9: Predictive Cache
204 
205**9.1 Check Status**:
206- `predictive_cache_status()` → shows if enabled
207 
208**9.2 Access Patterns**:
209- `access_patterns(limit=5)` → learned patterns
210 
211**9.3 Predict Next** (needs existing access history):
212- `predict_next(memory_id)` → predicted memories
213 
214**9.4 Warm Cache**:
215- `warm_cache(memory_id)` → pre-promote predicted
216 
217---
218 
219### Phase 10: Retrieval Quality
220 
221**10.1 Mark Memory Used**:
222- `mark_memory_used(memory_id, feedback="helpful")`
223 
224**10.2 Retrieval Stats**:
225- `retrieval_quality_stats()` → global stats
226- `retrieval_quality_stats(memory_id=X)` → per-memory
227 
228---
229 
230### Phase 11: Maintenance & DB Info
231 
232**11.1 Stats & Observability**:
233- `memory_stats()`
234- `hot_cache_status()`
235- `metrics_status()`
236 
237**11.2 Database Info**:
238- `db_info()` → schema version, size
239- `embedding_info()` → provider, cache info
240 
241**11.3 Maintenance Operations**:
242- `db_maintenance()`
243- `validate_embeddings()`
244- `run_cleanup()` → comprehensive cleanup
245 
246**11.4 Consolidation**:
247- `preview_consolidation()` → dry run
248- `run_consolidation(dry_run=true)` → preview
249- `run_consolidation(dry_run=false)` → actual merge (careful!)
250 
251**11.5 Audit History**:
252- `audit_history(limit=10)`
253 
254**11.6 Vector Rebuild** (use with caution - rebuilds all embeddings):
255- `db_rebuild_vectors(batch_size=100)` → re-embed all memories
256- Use when: switching embedding models, fixing dimension mismatches, recovering from corruption
257 
258---
259 
260### Phase 12: MCP Resources
261 
262**12.1 Hot Cache Resource**:
263Read `memory://hot-cache` directly (auto-injected to Claude):
264- Contains all promoted memories for instant recall
265- Verify contents match `hot_cache_status()` items
266 
267**12.2 Working Set Resource**:
268Read `memory://working-set` directly:
269- Session-aware active context (~10 items max)
270- Combines: recently recalled, predicted next, top salience
271- Verify smaller/more focused than hot-cache
272 
273**12.3 Project Context Resource**:
274Read `memory://project-context` directly:
275- Project-scoped memories for current working directory
276- Should filter to current project only
277 
278---
279 
280### Phase 13: Additional Tools
281 
282**13.1 List Memories**:
283- `list_memories(limit=5)` → paginated browse
284- `list_memories(memory_type="pattern")` → filtered
285- `list_memories(offset=5, limit=5)` → pagination
286 
287**13.2 Recall with Fallback**:
288- `recall_with_fallback("query")` → tries patterns → project → all
289 
290**13.3 Relationship Stats**:
291- `relationship_stats()` → knowledge graph overview
292 
293**13.4 Session Details**:
294- `get_session(session_id)` → specific session
295- `get_session_memories(session_id)` → memories from session
296- `cross_session_patterns()` → patterns across sessions
297 
298---
299 
300### Phase 14: Error Handling & Edge Cases
301 
302**14.1 Invalid IDs**:
303- `forget(memory_id=999999)` → should return error/not found
304- `promote(memory_id=-1)` → should handle gracefully
305- `get_related_memories(memory_id=0)` → should not crash
306 
307**14.2 Recall Edge Cases**:
308- `recall("", mode="precision")` → empty query handling
309- `recall("xyz", threshold=0.99)` → very high threshold, likely no results
310- `recall("test", limit=0)` → zero limit edge case
311- `recall("test", limit=1000)` → large limit handling
312 
313**14.3 Pagination Boundaries**:
314- `list_memories(offset=99999, limit=10)` → beyond data range
315- `list_memories(offset=-1)` → negative offset
316- `audit_history(limit=0)` → zero limit
317 
318**14.4 Link/Unlink Errors**:
319- `link_memories(id, id, "relates_to")` → self-link
320- `link_memories(999, 888, "relates_to")` → non-existent IDs
321- `unlink_memories(id_a, id_b)` → when no link exists
322 
323**14.5 Trust Boundaries**:
324- `validate_memory(id, boost=10.0)` → extreme boost (should cap at 1.0)
325- `invalidate_memory(id, penalty=10.0)` → extreme penalty (should floor at 0.0)
326 
327**14.6 Session Errors**:
328- `get_session("nonexistent-session-id")` → invalid session
329- `end_session("bad-id")` → non-existent session
330 
331**14.7 Mining Edge Cases**:
332- `run_mining(hours=0)` → zero hours
333- `approve_candidate(pattern_id=999)` → non-existent candidate
334 
335---
336 
337### Phase 15: Cleanup
338 
339After testing, clean up test data:
340- `forget()` all memories tagged with "test"
341- Or use `recall_by_tag("test")` to find them first
342 
343---
344 
345### Phase 16: Compact Conversation
346 
347**16.1 Run Compact**:
348After completing all tests, prompt the user to run `/compact` to:
349- Reduce conversation context size
350- Test that the session can be resumed after compaction
351- Verify memories persist across conversation summarization
352 
353Note: `/compact` is a user-initiated command and cannot be run programmatically by the assistant.
354 
355---
356 
357## Test Tracking
358 
359Track results as you go:
360 
361| Phase | Test | Status | Notes |
362|-------|------|--------|-------|
363| 1.1 | Remember | ⬜ | IDs: |
364| 1.2 | Recall | ⬜ | |
365| 1.3 | Recall Modes | ⬜ | |
366| 1.4 | Recall by Tag | ⬜ | |
367| 1.5 | Forget | ⬜ | |
368| 2.1 | Promotion | ⬜ | |
369| 2.2 | Demotion | ⬜ | |
370| 2.3 | Pin/Unpin | ⬜ | |
371| 3.1 | Link Memories | ⬜ | IDs: A=, B=, C= |
372| 3.2 | Get Related | ⬜ | |
373| 3.3 | Multi-Hop | ⬜ | |
374| 3.4 | Unlink | ⬜ | |
375| 4.1 | Validate | ⬜ | |
376| 4.2 | Invalidate | ⬜ | |
377| 4.3 | Trust History | ⬜ | |
378| 5.1 | Contradictions | ⬜ | IDs: X=, Y= |
379| 5.2 | Mark/Find | ⬜ | |
380| 5.3 | Resolve | ⬜ | |
381| 6.1 | Sessions | ⬜ | |
382| 6.2 | Episodic | ⬜ | |
383| 6.3 | Session Topic | ⬜ | |
384| 6.4 | Summarize Session | ⬜ | |
385| 7.1 | Log Output | ⬜ | |
386| 7.2 | Run Mining | ⬜ | |
387| 7.3 | Review | ⬜ | |
388| 8.1 | Seed from Text | ⬜ | |
389| 8.2 | Seed from File | ⬜ | |
390| 8.3 | Bootstrap | ⬜ | |
391| 9.1 | Predictive Status | ⬜ | |
392| 9.2 | Access Patterns | ⬜ | |
393| 9.3 | Predict Next | ⬜ | |
394| 9.4 | Warm Cache | ⬜ | |
395| 10.1 | Mark Used | ⬜ | |
396| 10.2 | Retrieval Stats | ⬜ | |
397| 11.1 | Stats | ⬜ | |
398| 11.2 | DB Info | ⬜ | |
399| 11.3 | Maintenance | ⬜ | |
400| 11.4 | Consolidation | ⬜ | |
401| 11.5 | Audit | ⬜ | |
402| 11.6 | Vector Rebuild | ⬜ | |
403| 12.1 | Hot Cache Resource | ⬜ | |
404| 12.2 | Working Set Resource | ⬜ | |
405| 12.3 | Project Context Resource | ⬜ | |
406| 13.1 | List Memories | ⬜ | |
407| 13.2 | Recall Fallback | ⬜ | |
408| 13.3 | Relationship Stats | ⬜ | |
409| 13.4 | Session Details | ⬜ | |
410| 14.1 | Invalid IDs | ⬜ | |
411| 14.2 | Recall Edge Cases | ⬜ | |
412| 14.3 | Pagination Boundaries | ⬜ | |
413| 14.4 | Link/Unlink Errors | ⬜ | |
414| 14.5 | Trust Boundaries | ⬜ | |
415| 14.6 | Session Errors | ⬜ | |
416| 14.7 | Mining Edge Cases | ⬜ | |
417| 15 | Cleanup | ⬜ | |
418| 16.1 | Compact Conversation | ⬜ | |
419 
420## Quick Smoke Test
421 
422For a 2-minute sanity check:
4231. `remember("Smoke test", tags=["smoke"])`
4242. `recall("smoke")`
4253. `promote(id)` → `hot_cache_status()`
4264. `demote(id)` → `forget(id)`
4275. `memory_stats()`
428 
429---
430 
431## Execution Mode
432 
433When running this skill:
4341. Ask user which phase to start with (or start from Phase 1)
4352. Execute each test, showing tool calls and results
4363. Update the tracking table after each test
4374. Pause between phases to let user review
4385. Note any failures or unexpected behavior
4396. Offer to skip phases or run specific tests
440