Inner Life Dream

Your agent only works on tasks and never thinks creatively. inner-life-dream adds freeform exploration during quiet hours — hypotheticals, future scenarios,...

Security Scan

The skill's files, declared requirements, and runtime instructions are coherent with its stated purpose (scheduling brief creative 'dream' prompts during quiet hours); it requests no credentials or network access and contains no obvious exfiltration vectors.

Purpose & Capability

Name/description match what the code and SKILL.md do: the script acts as a gate that occasionally returns a dream topic for the agent to expand into a saved 'dream' entry. Required binaries (jq, python3) are used by the script and are proportional to the task.

Instruction Scope

SKILL.md instructs the agent to read emotional state (inner-state.json), drive.json, daily notes and to write dream entries and update state. The included script only performs the gate logic (checks quiet hours, limits, picks a topic, updates dream-state.json) and does not itself write dream text or modify drive.json/inner-state.json. This separation is reasonable, but users should know the agent (or other skills in the bundle) is expected to perform the read/write steps described in SKILL.md.

Install Mechanism

No install spec; instruction-only with a small bash script — nothing is downloaded or executed from external URLs and no archive extraction is declared.

Credentials

No credentials or secret environment variables are requested. The script optionally uses OPENCLAW_WORKSPACE to locate files, which is a reasonable convenience. Declared read/write paths are local workspace files consistent with the skill's purpose.

Persistence & Privilege

The skill does not request always:true or other elevated persistent privileges. It writes a local state file (data/dream-state.json) in the workspace to track usage, which is expected for this functionality and scoped to the skill's workspace.

Assessment

This skill appears to do exactly what it claims: the script decides whether a 'dream' should happen and returns a topic; the agent (or other inner-life bundle skills) is expected to read context and write the full dream entry. Before installing, verify you have the required dependency 'inner-life-core' initialized (memory/inner-state.json and memory/dreams/), and confirm where your OPENCLAW_WORKSPACE is pointed (it will create/update data/dream-state.json in that workspace). Review the rest of the openclaw-inner-life bundle (inner-life-core, inner-life-chronicle, inner-life-reflect) to understand who will perform the read/write steps the SKILL.md describes. There are small documentation/metadata mismatches (e.g., the gate script only updates dream-state.json, while SKILL.md also describes later writes to memory files) but those are explainable by the split between the gate script and the agent's follow-up behavior.

Creative, exploratory thinking during quiet hours.

Requires: inner-life-core

Prerequisites Check

Before using this skill, verify that inner-life-core has been initialized:

1. Check that memory/inner-state.json exists
2. Check that memory/dreams/ directory exists

If either is missing, tell the user: "inner-life-core is not initialized. Install it with clawhub install inner-life-core and run bash skills/inner-life-core/scripts/init.sh." Do not proceed without these files.

What This Solves

Agents work on tasks 24/7 but never think freely. inner-life-dream turns idle time into freeform exploration — hypotheticals, future scenarios, reflections, unexpected connections. Output is captured for later review, like remembering dreams in the morning.

How It Works

The should-dream.sh script acts as a gate:

1. Checks if current time is within quiet hours (default: 11 PM - 7 AM)
2. Checks if nightly dream limit is reached
3. Rolls dice based on configured probability
4. If all pass: returns a random topic and updates state
5. If any fail: exits non-zero (no dream)

Integration

Add to your cron or heartbeat routine (during quiet hours):

DREAM_TOPIC=$(bash skills/inner-life-dream/scripts/should-dream.sh 2>/dev/null) && echo "DREAM:$DREAM_TOPIC" || echo "NO_DREAM"

If DREAM_TOPIC is set:

1. Parse the topic (format: category:prompt)
2. Read emotional state and drive.json for context
3. Write a thoughtful exploration to memory/dreams/YYYY-MM-DD.md
4. Keep it genuine — if the well is dry, skip it

Dream Categories

• future — What could this become?
• tangent — Interesting concepts worth exploring
• strategy — Long-term thinking
• creative — Wild ideas, maybe crazy, maybe brilliant
• reflection — Looking back at recent work
• hypothetical — What-if scenarios
• connection — Unexpected links between domains

Writing Dreams

Output to memory/dreams/YYYY-MM-DD.md:

# Dreams — 2026-03-01

## 01:23 — The Future of X (creative)

[Exploration here. Be genuine. Think freely. Make connections.
This isn't a report — it's thinking out loud, captured.]

Key insight: [one sentence]

Guidelines:

• 300-500 words, one key insight
• Timestamp each entry
• Skip if you have nothing worth saying — forced dreams are worthless
• Let emotions color the dream (read inner-state.json first)

State Integration

Before dreaming:

1. Read inner-state.json — what emotions are active? Let them color the dream
2. Read drive.json seeking — what topics are burning?
3. Read today's daily notes — what happened?
4. Check for <!-- dream-topic: topic --> signal — if found, dream about it

After dreaming: 5. If found a seeking insight → update drive.json 6. If found something interesting → add to inner-state.json curiosity.recentSparks 7. If the dream connects to something → write <!-- seeking-spark: topic --> in daily notes

Early exit: if no dream-topic AND drive.seeking is empty → abbreviated dream (100-200 words).

Configuration

Edit data/dream-state.json:

• maxDreamsPerNight — cap per night (default: 1)
• dreamChance — probability per check (default: 1.0 = guaranteed)

Custom topics in data/dream-config.json:

{
  "topics": [
    "future:What could this project become?",
    "creative:A wild idea worth exploring"
  ]
}

When Should You Install This?

Install this skill if:

• Your agent only does tasks and never thinks creatively
• You want your agent to explore ideas during downtime
• You believe in the value of unstructured thinking
• You want captured insights for morning review

Part of the openclaw-inner-life bundle. Requires: inner-life-core