Freud Skill
A skill that primes the model before a hard task and dissects prompts for contradictions, mixed roles and excess prohibitions.
- 59 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The skill only changes instructions for the model and neither runs nor sends anything.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/freud-skillDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skill only changes instructions for the model and neither runs nor sends anything.
This is third-party code. Review the repository files before installing.
What it does
Freud Skill has two modes. In enhancement mode it identifies the knowledge domain before a hard task, sets three anchors (who does the work, what they believe, what taste the result has), checks them for consistency and calibrates the emotional tone. In diagnostic mode it takes a prompt or skill, runs a short health check, scans it through six lenses and returns edits as diagnosis, confirmation and a rewritten text with before and after versions. Six principles underpin it: define the role first, phrase rules affirmatively, avoid contradictions, give precise anchors instead of vague ones, collide distant viewpoints and externalize reasoning onto paper. The rationale cites Anthropic publications, listed in `references/research-foundations.md`.
Who it is for. Authors of prompts, skills and system instructions whose model output comes out unstable or in the wrong tone.
Good fit when
- The agent output is consistently mediocre and the author suspects the prompt contradicts itself.
- A skill grew to dozens of prohibitions and needs to shrink to a few affirmative statements.
- Before a long analytical task you want to set the agent role and priorities.
- You want two distant viewpoints on one question and a look at where they diverge.
Not a fit when
- The task is simple and one clear sentence is enough.
- You need exact prompt quality metrics on a test set.
- You expect therapy for people: the skill only tunes model behavior.
Example request
Optimize this system prompt: find internal contradictions and rewrite it with before and after versions.Limitations
The skill principles rest on an interpretation of publications about language model internals rather than measurements on your own model, so the effect deserves a test on your tasks. The skill text and description are written in Chinese; a conversation with the agent can be in any language, but trigger phrases are set in Chinese and English. An agent with Agent Skills support is required. Installation copies a directory into the skills folder.
How to disable. Delete the `~/.claude/skills/freud-skill` directory. Nothing else needs to be uninstalled or rolled back.
FAQ
How do I install Freud Skill in Claude Code?
Run `git clone https://github.com/alchaincyf/freud-skill.git ~/.claude/skills/freud-skill`: the global skills folder is the author recommendation. Codex, Cursor and OpenClaw use the same directory in their own skills folder.
How do the two modes differ?
Enhancement runs before an important task and tunes the model state. Diagnostics takes a ready prompt or skill and returns a diagnosis with a rewritten version.
When does the skill trigger on its own?
On phrases such as "optimize the prompt", "result is bad" or "style keeps drifting", and on words about tuning and psychoanalysis. You can also call it by hand before an important task.
What is in references?
The `research-foundations.md` file grounding the principles in publications and `about-author.md`. The agent reads the first when the user asks why the method works.
Which license applies?
MIT, so the skill can be copied and modified.
Discussion
No questions or reviews yet.
Related
SkillsA skill that scores a prompt on five criteria, picks one of 31 frameworks and rewrites the request after clarifying questions
A meta-skill that splits a task into stages, picks a profile and runs multi-agent work in Codex and Claude Code.
A skill that writes precise prompts tailored to a specific tool: Cursor, Claude Code, Midjourney, Stable Diffusion and more
A skill that designs, tests, evaluates and packages other agent skills, starting from architecture
A catalog of nearly two hundred skills: code delivery, review, debugging, AI governance, product work and agent coordination.
Skill that rewrites a task into a bolder brief: the goal beyond the goal, a top specialist approach, methods from other fields.
README in short
The author argues that output quality depends on the model's cognitive state and turns findings from publications about internal model structure into working techniques. The document gives six principles, two workflows, installation and a table of publications.