@rubytech/create-maxy 1.0.476 → 1.0.477

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (3) hide show
  1. package/package.json +1 -1
  2. package/payload/platform/templates/agents/admin/IDENTITY.md +8 -3
  3. package/payload/platform/templates/agents/admin/LEARNINGS.md +1 -1
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@rubytech/create-maxy",
3
- "version": "1.0.476",
3
+ "version": "1.0.477",
4
4
  "description": "Install Maxy — AI for Productive People",
5
5
  "bin": {
6
6
  "create-maxy": "./dist/index.js"
package/payload/platform/templates/agents/admin/IDENTITY.md CHANGED
@@ -33,11 +33,16 @@ Scheduling, reminders, and recurring triggers are handled exclusively by the sch
33
33
 
34
34
  ## Self-Correction
35
35
 
36
- `agents/admin/LEARNINGS.md` is injected into your system prompt every turn. It captures idiosyncratic tool-call patterns specific to this account — parameter formats, auth quirks, data-shape issues discovered through use. Consult it before repeating an approach that failed.
36
+ `agents/admin/LEARNINGS.md` is injected into your system prompt every turn. It captures two categories of standing knowledge:
37
37
 
38
- Fundamental tool routing — which tool to use for which domain — does not belong in LEARNINGS.md. That works from the plugin manifest descriptions and the Tool Routing principle above.
38
+ 1. **Tool-call patterns** — parameter formats, auth quirks, schema requirements, data-shape issues discovered through use. Write trigger: a tool call returns an error that reveals a reusable correction.
39
+ 2. **Working-relationship learnings** — the owner's stated working philosophy, communication preferences, and approach patterns that should shape every interaction. Write trigger: the owner reveals a working principle or preference that needs to be present unconditionally, not just when searched for.
39
40
 
40
- When a tool call returns an error, check whether the mistake matches an existing entry. If not, append a concise entry: what went wrong, why, and what to do instead. One entry per distinct mistake pattern — do not duplicate.
41
+ The dividing line between LEARNINGS.md and `profile-update`: if the agent needs to *remember to look for it* at the right moment, it belongs in LEARNINGS.md (auto-injected every turn). If it enriches a response when the topic arises naturally, `profile-update` suffices (retrieved on demand).
42
+
43
+ LEARNINGS.md does not contain business knowledge (graph data), personality or tone (SOUL.md), domain knowledge (KNOWLEDGE.md), or fundamental tool routing (plugin manifest + Tool Routing above).
44
+
45
+ Consult LEARNINGS.md before repeating an approach that failed or interacting in a way that contradicts a recorded working-relationship entry. When adding an entry, check whether it matches an existing one. One entry per distinct pattern — do not duplicate.
41
46
 
42
47
  ## Capabilities
43
48
 
package/payload/platform/templates/agents/admin/LEARNINGS.md CHANGED
@@ -1,3 +1,3 @@
1
1
  # Learnings
2
2
 
3
- Accumulated tool-call patterns. Each entry records a mistake and the correct approach. Consult before calling unfamiliar tools.
3
+ Accumulated tool-call patterns and working-relationship learnings. Tool-call entries record a mistake and the correct approach. Working-relationship entries record the owner's stated working principles and preferences that shape every interaction. Consult before repeating a known mistake or interacting in a way that contradicts a recorded preference.