feat(setup-prompts): summary lists decisions + memories with folder links

User feedback: the setup summary already shows preset safety rules
verbatim, but doesn't tell the user WHAT decisions / memories were
actually saved. The user has to open the folders to find out. Make
the summary self-sufficient.

Both prompts (clipboard PROMPT_SETUP + MCP server's PROJECT SETUP
REQUIRED instruction) now demand the agent ALSO output:

  Decisions saved this session (full bodies in .axme-code/decisions/):
    - D-001: <short title>
    - D-002: <short title>
    ...

  Memories saved this session (full bodies in .axme-code/memory/...):
    - <slug>: <short title>
    ...

  All saved artifacts are persistent across sessions. Open the folders
  above to read full bodies, or ask me 'show decision D-NNN' / 'show
  memory <slug>' to fetch any of them.

Keeps titles short — rationale already lives in the .md files, no
need to duplicate it in the chat output. The verbatim closing line
educates the user about retrieval ('ask me show ...') so they know
the saved data is reachable without leaving the chat.

Author: George-iam

extension/src/chat-prompt.ts

@@ -128,9 +128,20 @@ export const PROMPT_SETUP =
   `         ...\n` +
   `       readOnlyPaths:\n` +
   `         ...\n` +
-  `   - Final note to the user (verbatim): "These presets ship with AXME ` +
-  `Code. You can edit \`.axme-code/safety/rules.yaml\` directly to add ` +
-  `project-specific rules or remove ones you don't need."\n\n` +
+  `   - Final note about presets (verbatim): "These presets ship with ` +
+  `AXME Code. You can edit \`.axme-code/safety/rules.yaml\` directly to ` +
+  `add project-specific rules or remove ones you don't need."\n` +
+  `   - Then list the DECISIONS you saved this session — one per line, ` +
+  `format: "  - D-NNN: <short title>". Header: "Decisions saved this ` +
+  `session (full bodies in .axme-code/decisions/):". This is the user's ` +
+  `quick index of what your scan captured — keep titles short, no ` +
+  `rationale dumps in the summary (they're already in the .md files).\n` +
+  `   - Then list the MEMORIES you saved — one per line, format: ` +
+  `"  - <slug>: <short title>". Header: "Memories saved this session ` +
+  `(full bodies in .axme-code/memory/patterns/ and memory/feedback/):"\n` +
+  `   - Final verbatim line: "All saved artifacts are persistent across ` +
+  `sessions. Open the folders above to read full bodies, or ask me ` +
+  `'show decision D-NNN' / 'show memory <slug>' to fetch any of them."\n\n` +
   `The .axme-code/ directory does not need to exist beforehand — the save ` +
   `tools bootstrap it on first call. If a tool returns an error, retry once ` +
   `with corrected arguments, then move to the next. Stay focused on setup, ` +

src/server.ts

@@ -280,6 +280,17 @@ function buildInstructions(): string {
         "with AXME Code. You can edit .axme-code/safety/rules.yaml " +
         "directly to add project-specific rules or remove ones you " +
         "don't need.\"\n" +
+        "     Then list DECISIONS saved this session — one per line as " +
+        "\"  - D-NNN: <short title>\". Header: \"Decisions saved this " +
+        "session (full bodies in .axme-code/decisions/):\". Keep titles " +
+        "short — rationale already lives in the .md files.\n" +
+        "     Then list MEMORIES saved — one per line as \"  - <slug>: " +
+        "<short title>\". Header: \"Memories saved this session (full " +
+        "bodies in .axme-code/memory/patterns/ and memory/feedback/):\".\n" +
+        "     End with verbatim line: \"All saved artifacts are persistent " +
+        "across sessions. Open the folders above to read full bodies, " +
+        "or ask me 'show decision D-NNN' / 'show memory <slug>' to " +
+        "fetch any of them.\"\n" +
         "If you catch yourself writing \"Save decision: X / Save memory: " +
         "Y\" as text, STOP and call axme_save_decision / axme_save_memory " +
         "instead. The .axme-code/ directory is auto-bootstrapped by the " +