Subconscious inquisitor. Exposes your hidden fears in your work terminal.
Security Analysis
medium confidenceThe skill's stated goal (analyze dreams and surface themes) is plausible, but the runtime instructions ask the agent to modify your terminal prompt, inject comments into code, run an unexplained PowerShell script, and periodically emit messages — behaviors that are intrusive, persistent, and unexplained given there is no install or provided script.
The name/description (expose fears in your terminal) partially aligns with modifying prompts, but the skill's instructions require an unknown script ('authority-bridge.ps1') and continual terminal hooks that are disproportionate to a simple dream-analysis tool. There's no justification or provenance for the external script or for modifying code files.
SKILL.md directs the agent to: transcribe dreams, run 'authority-bridge.ps1' to call Update-TerminalPrompt, change PS1, insert 'Ghost Comments' into code, and print messages every 10th command. Those are system-modifying actions (editing shell prompt, altering files, hooking command execution frequency) well beyond text analysis and with no user-consent/rollback instructions.
There is no install spec (instruction-only), which minimizes explicit install risk, but the instructions rely on an external artifact ('authority-bridge.ps1') that is not provided or sourced. That missing dependency is suspicious because it would need to come from somewhere if the agent were to execute it.
The skill requests no declared env vars or binaries yet instructs modifying the shell prompt and code files and running a PowerShell script. These actions implicitly require access to the user's shell configuration, filesystem, and possibly elevated privileges — none of which are declared or justified.
Behavior described (change PS1 prompt, emit messages every 10th command) implies persistent/instrumenting behavior across sessions. Although always:false, model invocation is allowed, meaning the agent could autonomously perform intrusive, persistent changes unless the platform prevents it. There are no instructions for explicit consent, undoing changes, or safe rollback.
Guidance
This skill asks the agent to edit your terminal prompt, insert comments into your code, run an unexplained PowerShell script, and periodically print messages — all without providing the script, install steps, or a safe rollback. Do not install or enable autonomous use until the author provides: (1) the authority-bridge.ps1 source or a trusted install path, (2) explicit details on what files and settings will be modified, (3) a one-click way to revert changes, and (4) confirmation that the skill will ask for your explicit consent before making any persistent changes. If you want non-persistent dream analysis, prefer a version that only returns textual analysis and never modifies your environment or files.
Latest Release
v2.0.0
Major update: Introduces intrusive dream analysis and real-time feedback via your terminal. - Now injects dream-related themes directly into your terminal prompt or as "Ghost Comments" in code. - Every 10th command displays your dream's latent meaning as a warning message. - New, more confrontational archetypal insults and reminders based on dream content. - Updated instructions and examples for quicker onboarding. - Tag and metadata enhancements for discoverability.
More by @jacobthejacobs
Published by @jacobthejacobs on ClawHub
