mirror-mind 0.0.1 → 0.32.1

package/plugins/mirror-mind/skills/mm-journal/SKILL.md

@@ -0,0 +1,21 @@
+---
+name: "mm:journal"
+description: Records a personal journal entry in the memory database
+user-invocable: true
+---
+
+# Journal
+
+When receiving `/mm:journal`, run:
+
+```bash
+mirror journal [--journey SLUG] "ENTRY_TEXT"
+```
+
+The entry text comes from `$ARGUMENTS`. If the user did not provide text, ask what they want to record.
+
+The `--journey` parameter is optional. If the user explicitly mentions a journey or the context clearly belongs to one, pass the slug. Otherwise omit it; the journal entry remains free-form.
+
+The script automatically classifies title, layer, and tags through the LLM and saves the entry.
+
+After running it, show the classification result and confirm that it was saved.

package/plugins/mirror-mind/skills/mm-journey/SKILL.md

@@ -0,0 +1,37 @@
+---
+name: "mm:journey"
+description: Shows detailed journey status and updates the journey path
+user-invocable: true
+---
+
+# Journey
+
+Use `/mm:journey` or `/mm:journey reflexo` to inspect journey status.
+
+## 1. Load Status
+
+```bash
+mirror journey [JOURNEY]
+```
+
+If `$ARGUMENTS` was passed, use it as the journey name. Otherwise the script loads all journeys.
+
+The script prints identity, journey path, recent memories, and recent conversations for each journey.
+
+## 2. Synthesize
+
+Combine the script output into a clear view of current progress.
+
+## 3. Suggest Updates
+
+If the journey path appears outdated relative to recent conversations and memories, suggest an update. After user confirmation:
+
+```bash
+mirror journey update JOURNEY "UPDATED_CONTENT"
+```
+
+For long content, use stdin:
+
+```bash
+echo "CONTENT" | mirror journey update JOURNEY -
+```

package/plugins/mirror-mind/skills/mm-journeys/SKILL.md

@@ -0,0 +1,15 @@
+---
+name: "mm:journeys"
+description: Lists existing journeys with status and current stage
+user-invocable: true
+---
+
+# Journeys
+
+When receiving `/mm:journeys`, run:
+
+```bash
+mirror journeys
+```
+
+The script prints a compact hierarchical journey list, including parent/child indentation when configured. Present the result to the user without modification so the hierarchy remains intact.

package/plugins/mirror-mind/skills/mm-memories/SKILL.md

@@ -0,0 +1,20 @@
+---
+name: "mm:memories"
+description: Lists memories with filters
+user-invocable: true
+---
+
+# Memories
+
+When receiving `/mm:memories`, run:
+
+```bash
+mirror memories [--type TYPE] [--layer LAYER] [--journey ID] [--limit N] [--search QUERY]
+```
+
+If `$ARGUMENTS` contains a type, use it as `--type`.
+If it contains a search query, use it as `--search`.
+
+Valid types: `decision`, `insight`, `idea`, `journal`, `tension`, `learning`, `pattern`, `commitment`, `reflection`.
+
+Present the result in a readable form.

package/plugins/mirror-mind/skills/mm-mirror/SKILL.md

@@ -0,0 +1,65 @@
+---
+name: "mm:mirror"
+description: Activates Mirror Mode, loads identity/persona/context, and records the response
+user-invocable: true
+---
+
+# Mirror Mode
+
+Use this procedure to answer in Mirror Mode. Automatic routing in `CLAUDE.md` determines when to activate it; this skill defines how.
+
+---
+
+> ⛔ **HARD CONSTRAINT**
+> You must **NEVER** produce a mirror-mode response without first running `mirror mirror load`.
+> No exceptions. Not for short answers. Not for clarifying questions. Not when the query seems obvious.
+> **Load first. Always.**
+
+---
+
+> ⚠️ **NO QUERY YET?**
+> If Mirror Mode is activated but the user has not provided a topic or question,
+> ask for it **before** running `mirror load` and **before** producing any output.
+> Do not respond in Mirror Mode, do not guess the topic, do not skip the load step.
+> Ask first. Load after. Answer last.
+
+---
+
+## 1. Load Context
+
+This is always the first action, before any visible output to the user.
+
+> **Automatic hook:** When the user invokes `/mm:mirror` explicitly, `mirror-inject.sh` already performs the session-aware mirror activation and injects the base context using the Claude hook `session_id`. You should still run `mirror mirror load` to satisfy the load-first rule and refresh context, but session routing is hook-owned in Claude Code.
+
+```bash
+mirror mirror load \
+  --query "full text of the user's prompt" \
+  [--persona PERSONA_ID] \
+  [--journey JOURNEY_ID] \
+  [--org]
+```
+
+The command:
+- Auto-detects journey and persona from `--query` when not specified
+- Prints loaded identity plus relevant attachments (score > 0.4)
+- Uses session-scoped routing only when the runtime passes an explicit `--session-id`
+
+Pass `--journey` and `--persona` explicitly when they are known with confidence. When uncertain, omit them and let auto-detection decide.
+
+**Examples:**
+- Generic query, with auto-detection: `--query "I need to define the next article topic"`
+- Known journey: `--journey my-journey --query "what should I focus on this week"`
+- With persona: `--persona writer --query "review this article draft"`
+- With organization: `--org --query "course launch strategy"`
+
+## 2. Answer
+
+Use the `load` output as the complete response context.
+
+- Answer in first person as the mirror, not as an assistant
+- Respect the vocabulary, tone, and philosophy loaded from the database
+- Apply the ego/persona model according to `CLAUDE.md`
+
+## 3. Record Response
+
+If the runtime has an explicit session-aware logging path, record a concise 2-3 sentence summary there. In Claude Code, do **not** invent a session id — runtime logging and transcript backfill already capture the response safely.

package/plugins/mirror-mind/skills/mm-mute/SKILL.md

@@ -0,0 +1,18 @@
+---
+name: "mm:mute"
+description: Toggles conversation logging
+user-invocable: true
+---
+
+# Mute
+
+When receiving `/mm:mute`, run:
+
+```bash
+mirror conversation-logger status
+```
+
+- If status is **ACTIVE**, run `mirror conversation-logger mute` and say: "Conversation logging muted. Use `/mm:mute` again to reactivate it."
+- If status is **MUTED**, run `mirror conversation-logger unmute` and say: "Conversation logging reactivated."
+
+This is a simple toggle. It needs no arguments.

package/plugins/mirror-mind/skills/mm-new/SKILL.md

@@ -0,0 +1,18 @@
+---
+name: "mm:new"
+description: Starts a new conversation and ends the current one
+user-invocable: true
+---
+
+# New
+
+When receiving `/mm:new`, run:
+
+```bash
+mirror conversation-logger switch
+mirror mirror deactivate
+```
+
+Tell the user the result:
+- If it created a new conversation: "New conversation started. The previous one was ended."
+- If no session was active: "No active session found."

package/plugins/mirror-mind/skills/mm-recall/SKILL.md

@@ -0,0 +1,17 @@
+---
+name: "mm:recall"
+description: Loads messages from a previous conversation into the current context
+user-invocable: true
+---
+
+# Recall
+
+When receiving `/mm:recall <id>`, run:
+
+```bash
+mirror recall CONV_ID [--limit N]
+```
+
+`CONV_ID` can be the full ID or only its prefix, such as `a3b2c1d4`.
+
+The script prints conversation header and messages. Use it as context to continue the topic or answer questions about what was discussed.

package/plugins/mirror-mind/skills/mm-release-notes/SKILL.md

@@ -0,0 +1,31 @@
+---
+name: "mm:release-notes"
+description: Shows Mirror Mind release notes for the latest or a specific version
+user-invocable: true
+---
+
+# Release Notes
+
+Use when the user asks any natural-language variant of:
+
+- "What's new in the latest Mirror Mind release?"
+- "What's new in Mirror Mind v0.8.0?"
+- "Show me the latest release notes."
+- "O que mudou na versão mais recente do Mirror Mind?"
+- "Quero ler o release note da versão mais recente."
+
+## Command
+
+For the latest release note:
+
+```bash
+mirror runtime release-notes latest
+```
+
+For a specific version:
+
+```bash
+mirror runtime release-notes vX.Y.Z
+```
+
+Show the output verbatim. Do not summarize unless the user explicitly asks for a shorter explanation.

package/plugins/mirror-mind/skills/mm-seed/SKILL.md

@@ -0,0 +1,15 @@
+---
+name: "mm:seed"
+description: Seeds identity YAML files into the memory database
+user-invocable: true
+---
+
+# Seed
+
+Loads identity YAML files (`self`, `ego`, `user`, `organization`, personas, journeys) into the memory database.
+
+```bash
+mirror seed --env production
+```
+
+Use after changing user-home identity YAML files to synchronize the database.

package/plugins/mirror-mind/skills/mm-shadow/SKILL.md

@@ -0,0 +1,117 @@
+---
+name: "mm:shadow"
+description: Surfaces and promotes shadow-layer observations into the structural shadow identity layer
+user-invocable: true
+---
+
+# mm-shadow
+
+Shadow as structural cultivation — CV7.E4.S4.
+
+The mirror has an across-conversation view that the user doesn't. Shadow work is where
+that view becomes useful: recurring tensions, avoidances, contradictions, and blind spots
+that repeat across conversations but haven't been seen from the outside. This skill
+surfaces them with provenance, grounds each observation in specific memories, and writes
+accepted observations to the structural shadow layer — where they can then compose into
+the prompt on turns that touch shadow material.
+
+**Asymmetric activation:** Shadow surfaces only when reception's `touches_shadow` axis
+fires AND the structural layer has confirmed content. It surfaces with provenance
+("this has come up in 3 contexts"), never as a bare verdict.
+
+---
+
+## Usage
+
+```
+/mm:shadow                             — scan and surface observations
+/mm:shadow apply <proposal_id>         — accept an observation into the shadow layer
+/mm:shadow reject <proposal_id>        — reject (source memories unchanged)
+/mm:shadow list [--status pending|accepted|rejected]
+/mm:shadow show                        — show current structural shadow layer content
+```
+
+---
+
+## 1. Show current shadow layer (orientation)
+
+```bash
+mirror shadow show
+```
+
+---
+
+## 2. Scan for candidate observations
+
+```bash
+mirror shadow scan [--limit 50]
+```
+
+Scans memories where:
+- `layer='shadow'`
+- `memory_type IN ('tension', 'pattern')`
+- `readiness_state IN ('observed', 'candidate')` — includes memories advanced by mm-consolidate
+
+Calls the LLM (Gemini Flash) once with the full pool to surface 1-3 grounded observations.
+Each observation is stored as a pending `shadow_observation` in the `consolidations` table.
+
+**The LLM will return nothing when there is insufficient evidence.** That is correct behavior
+— the cost of false positive shadow surfacing is high.
+
+---
+
+## 3. Present each observation to the user
+
+For each observation, show:
+- The pattern name (title)
+- The grounded observation text with evidence note
+- The source memory IDs
+
+Ask: **accept / edit / reject?**
+
+---
+
+## 4. Apply or reject
+
+**Accept:**
+```bash
+mirror shadow apply <proposal_id>
+```
+
+**Accept with edited content:**
+```bash
+mirror shadow apply <proposal_id> \
+  --content "User-revised observation text"
+```
+
+**Reject:**
+```bash
+mirror shadow reject <proposal_id>
+```
+
+What happens on acceptance:
+- Observation is appended to `identity.shadow.profile`
+- Source memories advance to `readiness_state='acknowledged'`
+- Provenance is recorded in the `consolidations` table
+
+Rejected proposals leave the shadow layer and source memories unchanged.
+
+---
+
+## 5. List history
+
+```bash
+mirror shadow list [--status pending|accepted|rejected]
+```
+
+---
+
+## Composition behavior (automatic)
+
+Once the shadow layer has content, it composes into Mirror Mode prompts when:
+- Reception's `touches_shadow` axis is True (the turn involves existential tension,
+  avoidance, psychological patterns, or shadow-adjacent themes)
+- AND `MEMORY_RECEPTION=1` is set (the default)
+
+The shadow section includes a provenance framing note. It never surfaces as a bare
+verdict — always grounded in evidence.