Dream Art

Transform a user's dream into actionable artwork concepts and generate outputs. Use when Uli describes a dream and wants to explore its creative potential —...

Security Scan

The skill's behavior mostly matches its description (transform dreams into creative outputs) but its runtime instructions expect and perform file-system writes via obsidian-cli and ask the agent to do a silent internal analysis — yet the skill metadata does not declare the obsidian dependency or make its read/write implications explicit.

Purpose & Capability

The name/description (convert dreams to creative outputs) aligns with the SKILL.md workflow (intake, analysis, generate). However, the instructions require use of the obsidian-cli tool to locate and append to a user's Obsidian daily note, yet the skill metadata lists no required binaries or config paths. That omission is an inconsistency: a skill that logs to an Obsidian vault reasonably should declare obsidian-cli (or describe vault path requirements).

Instruction Scope

The SKILL.md explicitly instructs the agent to run obsidian-cli to find the user's vault path and append the user's verbatim dream text into today's daily note. This means the skill will write user-provided content to the user's filesystem (Obsidian vault). It also directs the agent to perform a silent internal Phase 2 analysis ('Do not output this phase explicitly. Do it silently'), which grants the agent discretion to transform or retain user content without an explicit visible step. Both the file write and the 'silent' internal processing are within the skill's stated purpose but merit caution because the metadata does not disclose the file-write dependency and users may not expect silent analysis/storage.

Install Mechanism

This is an instruction-only skill with no install spec or code to download; that minimizes installation risk. The workflow assumes external tooling (obsidian-cli, and platform tools like image_generate/music_generate/video_generate) is available on the agent, but no packages are fetched by the skill itself.

Credentials

The skill requests no environment variables or credentials in metadata, which is proportionate. However, it does rely on access to the user's Obsidian vault via obsidian-cli, which implicitly requires filesystem access and permission to modify notes. That implicit requirement is not declared in requires.config or required binaries and should be called out to users.

Persistence & Privilege

The skill is not marked always:true and does not request persistent installation. It writes to the user's Obsidian vault (user data) but does not request platform-level persistent privileges. The ability to run autonomously is allowed by default (disable-model-invocation: false) but is not by itself a special red flag here.

What to consider before installing

This skill generally does what it says — turn dreams into creative outputs — but there are important practical and privacy details to consider before installing: - The SKILL.md instructs the agent to run obsidian-cli and append the user's verbatim dream text into today's Obsidian daily note. Verify you want your dreams stored in your vault and that obsidian-cli is installed and configured to point to the correct vault. - The skill metadata does not declare obsidian-cli as a required binary or the vault path as a required config, so treat that as a missing declaration rather than hidden behavior. Ask the author to document this dependency and what path will be written. - Phase 2 is performed 'silently' per the instructions. If you want full transparency, request that the skill surface its analysis before writing to files or make the logging step explicit and confirm before saving. - Confirm what 'image_generate', 'music_generate', and 'video_generate' tools actually do in your agent environment (where they send data or whether generation occurs locally) before sending sensitive or private dreams to them. If you plan to use the skill: ensure obsidian-cli is installed and points to a vault you control, review a sample run to confirm exactly what gets written, and ask the publisher to update metadata to declare the obsidian-cli dependency and the file-write behavior. If you cannot validate those, avoid installing or using the skill with sensitive content.

Dream Art — Skill

Transform raw dream content into creative output. Three-phase workflow.

---

Phase 1 — Intake & Log to Obsidian

Use obsidian-cli to log the dream to today's daily note.

1. Find today's daily note path:

   obsidian-cli print-default --path-only
   

   Daily notes are typically stored as Daily Notes/YYYY-MM-DD.md within the vault.

2. Append to today's daily note with structured metadata:

   ## Dream — [HH:MM]
   
   **Tags:** #dream #intake
   
   **Raw content:**
   [everything the user described — verbatim]
   
   **Emotional texture:** [1-2 sentences on the mood, feeling tone]
   **Key imagery:** [list the most vivid visual elements]
   **Notable symbols:** [recurring or charged objects/figures]
   **Unusual elements:** [things that don't follow ordinary logic]
   
   ---
   

3. Ask the user: "What medium feels right — image, music, animation, prose, or open?" If they say open, default to image generation.

---

Phase 2 — Dream Analysis (Internal)

Do not output this phase explicitly. Do it silently, then move to Phase 3.

Analyze the dream's creative potential:

• What is the dominant emotional frequency? (dread, longing, wonder, confusion, awe)
• What single image or moment is the "heart" of the dream?
• What medium best captures the texture — not just the content, but the feeling?
• What would an artwork need to include to make someone feel what the dream felt like?

Build toward a creative brief (use the ideation pattern if helpful):

• Named emotional intent
• Tone and negative space
• Reference points (if any come to mind)

---

Phase 3 — Generate & Deliver

Based on user's chosen medium:

Image: Use image_generate with a prompt derived from the dream's heart image. Include emotional texture in the prompt. Use image tool for reference if user has shared a style reference.

Music/Audio: Use music_generate with a prompt that captures the dream's emotional frequency. Ask user: "Song or ambient? Vocals or instrumental?"

Prose: Write a short piece (flash fiction, poem, scene) that transports the reader into the dream's world. 300-800 words.

Animation: Use video_generate for short atmospheric pieces (5-15s). Alternatively generate an image with strong motion implied.

After delivery: Ask "Want to push further — different medium, refine, or take it in a new direction?" (one question only)

---

Key Principles

• Log first, create second. The Obsidian capture makes the dream permanent and search-able.
• Emotional texture over literal content. Don't describe the dream back to the user — reframe it.
• One strong image beats ten weak ones. Find the heart.
• Never interpret the dream clinically. This is a creative tool, not a therapy session.
• Brief beats ramble. Phase 2 is for you — Phase 3 is what the user sees.