neural-memory

Name: neural-memory
Author: Rune-kit

80stars21forks

How to install neural-memory?

rune-kit/rune/neural-memory

$curl -o .claude/agents/neural-memory.md https://raw.githubusercontent.com/Rune-kit/rune/HEAD/agents/neural-memory.md

Installs into the current project.

Summary

Cross-session cognitive persistence via Neural Memory MCP. Captures decisions, patterns, errors, insights with semantic links. Bridges file-based state (session-bridge) with semantic graph.

Files · 1

View on GitHub

agents/neural-memory.md

1	# neural-memory
2
3	## Purpose
4
5	Bridges Rune's file-based persistence (session-bridge, journal) with Neural Memory MCP's semantic graph. While session-bridge saves decisions to `.rune/` files and journal tracks ADRs locally, neural-memory captures cross-project learnable patterns — decisions, error root causes, architectural insights, and workflow preferences — into a persistent cognitive layer that compounds across every project and session.
6
7	Without this skill, each project is an island. With it, a caching pattern discovered in Project A auto-surfaces when Project B faces a similar problem.
8
9	## Triggers
10
11	Auto-trigger:
12	- Session start → Run Recall Mode (load relevant context before any work)
13	- After `cook` completes a feature → Run Capture Mode (save learnings)
14	- After `debug` finds root cause → Run Capture Mode (save error pattern)
15	- After `review` finds issues → Run Capture Mode (save code quality insight)
16	- After `rescue` completes a phase → Run Capture Mode (save refactoring pattern)
17	- After `journal` writes an ADR → Run Capture Mode (extract to nmem)
18	- Session end / before compaction → Run Flush Mode (capture remaining context)
19
20	Manual trigger:
21	- `/rune recall <topic>` — search neural memory for a topic
22	- `/rune remember <text>` — save a specific memory
23	- `/rune brain-health` — check neural memory health + maintenance
24	- `/rune hypothesize <question>` — start hypothesis tracking
25
26	## Calls (outbound)
27
28	- `session-bridge` (L3): after Capture Mode — sync key decisions back to `.rune/` files
29
30	## Called By (inbound)
31
32	- `cook` (L1): Phase 0 (resume) + Phase 8 (complete) — recall context at start, capture learnings at end
33	- `rescue` (L1): phase start + phase end — recall past refactoring patterns, capture new ones
34	- `debug` (L2): after root cause found — capture error pattern for future recognition
35	- `fix` (L2): after fix verified — capture fix pattern (cause → solution)
36	- `review` (L2): after review complete — capture code quality insight
37	- `plan` (L2): before architecture decisions — recall past decisions on similar problems
38	- `sentinel` (L2): after security finding — capture vulnerability pattern
39	- `incident` (L2): after resolution — capture incident root cause + fix
40	- `retro` (L2): during retrospective — capture retro insights and patterns
41	- `session-bridge` (L3): Step 6 (cross-project extraction) — extract generalizable patterns
42	- `journal` (L3): after ADR written — extract decision + rejected alternatives
43	- `context-engine` (L3): before compaction — trigger Flush Mode to preserve context
44
45	## Modes
46
47	### Mode 1: Recall (Session Start / Before Decisions)
48
49	Load relevant context from neural memory before starting work.
50
51	Step 1 — Identify Recall Topics
52	Read `.rune/progress.md` and current task context to determine 3-5 diverse recall topics.
53	Always prefix queries with the project name to avoid cross-project noise.
54
55	```
56	GOOD: "Rune compiler cross-reference resolution"
57	GOOD: "MyTrend PocketBase auth session handling"
58	BAD: "cross-reference" (too generic, returns all projects)
59	BAD: "auth" (returns noise from every project)
60	```
61
62	Step 2 — Execute Recall
63	Call `nmem_recall` for each topic. Use diverse angles:
64	- Technology-specific: `"<project> React state management"`
65	- Problem-specific: `"<project> caching strategy decision"`
66	- Pattern-specific: `"<project> error handling approach"`
67
68	Step 3 — Synthesize Context
69	Summarize recalled memories into actionable context:
70	- Decisions that apply to current task
71	- Patterns that worked (or failed) before
72	- Constraints or preferences from past sessions
73	- Open hypotheses still being tracked
74
75	Step 4 — Surface Gaps
76	If recall returns thin results for the current domain, note the gap.
77	Call `nmem_gaps(action="detect")` if working in a domain with sparse memories.
78
79	---
80
81	### Mode 2: Capture (After Task Completion)
82
83	Extract learnable patterns from completed work and save to neural me

People who liked this also viewed

6 picks

Type

neural-memory

80stars21forks

View on GitHub

How to install neural-memory?

rune-kit/rune/neural-memory

$curl -o .claude/agents/neural-memory.md https://raw.githubusercontent.com/Rune-kit/rune/HEAD/agents/neural-memory.md

Installs into the current project.

Summary

Cross-session cognitive persistence via Neural Memory MCP. Captures decisions, patterns, errors, insights with semantic links. Bridges file-based state (session-bridge) with semantic graph.

Files · 1

View on GitHub

agents/neural-memory.md

1	# neural-memory
2
3	## Purpose
4
5	Bridges Rune's file-based persistence (session-bridge, journal) with Neural Memory MCP's semantic graph. While session-bridge saves decisions to `.rune/` files and journal tracks ADRs locally, neural-memory captures cross-project learnable patterns — decisions, error root causes, architectural insights, and workflow preferences — into a persistent cognitive layer that compounds across every project and session.
6
7	Without this skill, each project is an island. With it, a caching pattern discovered in Project A auto-surfaces when Project B faces a similar problem.
8
9	## Triggers
10
11	Auto-trigger:
12	- Session start → Run Recall Mode (load relevant context before any work)
13	- After `cook` completes a feature → Run Capture Mode (save learnings)
14	- After `debug` finds root cause → Run Capture Mode (save error pattern)
15	- After `review` finds issues → Run Capture Mode (save code quality insight)
16	- After `rescue` completes a phase → Run Capture Mode (save refactoring pattern)
17	- After `journal` writes an ADR → Run Capture Mode (extract to nmem)
18	- Session end / before compaction → Run Flush Mode (capture remaining context)
19
20	Manual trigger:
21	- `/rune recall <topic>` — search neural memory for a topic
22	- `/rune remember <text>` — save a specific memory
23	- `/rune brain-health` — check neural memory health + maintenance
24	- `/rune hypothesize <question>` — start hypothesis tracking
25
26	## Calls (outbound)
27
28	- `session-bridge` (L3): after Capture Mode — sync key decisions back to `.rune/` files
29
30	## Called By (inbound)
31
32	- `cook` (L1): Phase 0 (resume) + Phase 8 (complete) — recall context at start, capture learnings at end
33	- `rescue` (L1): phase start + phase end — recall past refactoring patterns, capture new ones
34	- `debug` (L2): after root cause found — capture error pattern for future recognition
35	- `fix` (L2): after fix verified — capture fix pattern (cause → solution)
36	- `review` (L2): after review complete — capture code quality insight
37	- `plan` (L2): before architecture decisions — recall past decisions on similar problems
38	- `sentinel` (L2): after security finding — capture vulnerability pattern
39	- `incident` (L2): after resolution — capture incident root cause + fix
40	- `retro` (L2): during retrospective — capture retro insights and patterns
41	- `session-bridge` (L3): Step 6 (cross-project extraction) — extract generalizable patterns
42	- `journal` (L3): after ADR written — extract decision + rejected alternatives
43	- `context-engine` (L3): before compaction — trigger Flush Mode to preserve context
44
45	## Modes
46
47	### Mode 1: Recall (Session Start / Before Decisions)
48
49	Load relevant context from neural memory before starting work.
50
51	Step 1 — Identify Recall Topics
52	Read `.rune/progress.md` and current task context to determine 3-5 diverse recall topics.
53	Always prefix queries with the project name to avoid cross-project noise.
54
55	```
56	GOOD: "Rune compiler cross-reference resolution"
57	GOOD: "MyTrend PocketBase auth session handling"
58	BAD: "cross-reference" (too generic, returns all projects)
59	BAD: "auth" (returns noise from every project)
60	```
61
62	Step 2 — Execute Recall
63	Call `nmem_recall` for each topic. Use diverse angles:
64	- Technology-specific: `"<project> React state management"`
65	- Problem-specific: `"<project> caching strategy decision"`
66	- Pattern-specific: `"<project> error handling approach"`
67
68	Step 3 — Synthesize Context
69	Summarize recalled memories into actionable context:
70	- Decisions that apply to current task
71	- Patterns that worked (or failed) before
72	- Constraints or preferences from past sessions
73	- Open hypotheses still being tracked
74
75	Step 4 — Surface Gaps
76	If recall returns thin results for the current domain, note the gap.
77	Call `nmem_gaps(action="detect")` if working in a domain with sparse memories.
78
79	---
80
81	### Mode 2: Capture (After Task Completion)
82
83	Extract learnable patterns from completed work and save to neural me

People who liked this also viewed

6 picks

Type