AI output is averaged, confident, and subtly wrong. This skill finds the specific thing that's off and tells you what to do about it.
A deslopification skill for Claude Code, Codex, Gemini, Cursor, Windsurf, and any LLM agent. Strip the averaged confidence, the universal claims, the polished nothing.
An LLM haunted by dead European philosophers, critiquing the output of LLMs that have never heard of them. It works anyway.
Input: AI-generated mission statement for an HR SaaS product.
To empower HR teams to run fairer, faster, and more meaningful performance
reviews by using AI to reduce administrative burden, improve consistency,
and help organizations invest in employee growth at scale.
"This is a solid mission statement. You might want to make it slightly more specific about what 'fairer' means in practice, and consider adding a concrete metric or outcome to strengthen the value proposition. The tone is professional and the focus on employee growth is a good differentiator."
"Fairer" is doing the most work and the least. The mission assumes AI makes reviews fairer by default — more consistent, less biased. But consistency applied to a flawed system just scales the flaw. The employee whose performance gets assessed on Slack response times and ticket velocity, who works fewer visible hours because they're a caregiver, who writes less polished prose because English isn't their first language — for them, faster and more consistent is worse, not better.
"Meaningful" has the same problem from the other direction. It assumes the purpose of a performance review is employee growth. For a significant portion of the organizations buying this, it's legal documentation for a termination, a compliance checkbox, or a PIP trail. AI that makes those processes faster isn't delivering meaning — it's scaling a mechanism that was already doing damage, now with better audit logs.
The sentence is selling a product to the person holding the gun by describing how much better the bullet will fly.
Revision: Rewrite from the employee's perspective, not the HR team's. "Help every employee understand where they stand and what comes next, with reviews that take hours instead of weeks." That's a mission. What's here is a feature list with the word "empower" in front.
| Lens | Theory | What It Does |
|---|---|---|
| 🛠️ Deconstruction | Peter Barry | Binary breakdown: find the oppositions, note the hierarchy |
| 🖼️ Structural Awareness | Patricia Waugh | See the frame around the text |
| 🚫 Grand Narrative | Jean-François Lyotard | Reject "one right way" thinking |
| 🌐 Hyperreal | Jean Baudrillard | Perfect content signals fake |
| 🌐 World-Building | Brian McHale | Pivot between realities |
| 🌓 Interpretive Agency | Roland Barthes | The AI's intent doesn't matter |
git clone https://github.com/kgeoffrey/postmodernist.git
cd postmodernistThe skill ships in two forms:
SKILL.md— Claude Code native format (references thereferences/folder for deep context)PROMPT.md— self-contained system prompt, works on any platform by paste or copy
The repo keeps human documentation at the root and keeps the postmodernist/ folder agent-facing only. That folder is the distributable skill package.
Global install (available in all projects):
mkdir -p "$HOME/.claude/skills"
cp -R postmodernist "$HOME/.claude/skills/"Project-level:
mkdir -p .claude/skills
cp -R postmodernist .claude/skills/Activate with:
Use the postmodernist skill on this AI-generated strategy document.
Native project-level skill install:
mkdir -p .agents/skills
cp -R postmodernist .agents/skills/Native personal install:
mkdir -p "$HOME/.agents/skills"
cp -R postmodernist "$HOME/.agents/skills/"If your Codex environment uses a different skills directory, copy the same postmodernist/ folder there.
Portable fallback:
cat postmodernist/PROMPT.md >> AGENTS.mdPrefer the native skill-folder install. Use the AGENTS.md fallback only when your Codex surface does not support skills yet.
Codex plugin package in this repo:
- Plugin root:
plugins/postmodernist/ - Repo marketplace entry:
.agents/plugins/marketplace.json
Project-level (GEMINI.md is read automatically from the project root):
cat postmodernist/PROMPT.md >> GEMINI.mdGlobal:
cat postmodernist/PROMPT.md >> "$HOME/.gemini/GEMINI.md"Add as a project rule:
mkdir -p .cursor/rules
cp postmodernist/PROMPT.md .cursor/rules/postmodernist.mdcOr globally via Cursor Settings → Rules → User Rules → paste PROMPT.md.
mkdir -p .windsurf/rules
cp postmodernist/PROMPT.md .windsurf/rules/postmodernist.mdmkdir -p .clinerules
cp postmodernist/PROMPT.md .clinerules/postmodernist.mdcat postmodernist/PROMPT.md >> .github/copilot-instructions.mdPaste the contents of PROMPT.md as your system prompt or custom instructions. It is fully self-contained — no external file references.
Claude / Codex / skill-aware agents:
Use postmodernist on this strategy document.
Apply postmodernist critique to this AI-generated mission statement.
Portable prompt mode:
Apply postmodernist critique to the AI-generated pitch.
Run a postmodernist audit on this strategy document.
Use the structural-awareness lens on this marketing copy.
| Mode | Trigger | What It Does |
|---|---|---|
| Default Critique | critique / audit / pressure-test |
Short continuous prose ending with a reframe and, for persuasive text, a concrete revision direction |
| Lens-Specific | use the structural-awareness lens on... |
Focus on one lens because the user explicitly asked for it |
| Structured Strategy Audit | give me a structured audit of this AI-generated strategy |
Narrative Audit, Authenticity Check, Frame Report, Revision Direction, and a final Frame Break |
Input: AI-generated product launch strategy
"Our sustainable tech product disrupts the market by delivering
eco-friendly innovation. Everyone wants sustainable options. Our
revolutionary platform approach ensures growth through
scale. The future is green."
Postmodernist Analysis:
"Everyone wants sustainable options" is the tell. The whole strategy
is built on a claim that collapses the moment you leave the audience
it was written for. Urban, $60k+, under 35: yes. Rural, price-first,
45+: no. "Revolutionary" and "disrupts" perform confidence the text
hasn't earned. The framing treats sustainable vs. unsustainable as
the buyer's choice — but buyers weigh value vs. cost, and the
strategy has nothing to say to the majority who land there.
The strategy is optimized for the audience it already has, not the
one it needs.
Revision: Lead with durability and cost-over-time. "Built to last
10 years, not 2." Fold sustainability in as proof for believers,
not as the headline for everyone.
.agents/
└── plugins/
└── marketplace.json # Repo-local Codex plugin marketplace entry
.github/
└── workflows/
└── validate.yml # CI for structural validation on pushes and PRs
plugins/
└── postmodernist/
├── .codex-plugin/plugin.json # Codex plugin manifest
└── skills/
└── postmodernist/ # Embedded skill copy for plugin distribution
postmodernist/
├── SKILL.md # Claude Code native skill
├── PROMPT.md # Self-contained prompt — paste on any platform
├── agents/
│ └── openai.yaml # Agent metadata
└── references/
├── 01-deconstruction-toolkit.md
├── 02-structural-awareness.md
├── 03-skepticism-grand-narratives.md
├── 04-hyperreal-navigation.md
├── 05-world-building.md
├── 06-interpretive-agency.md
└── ai-strategy-auditor.md # Full use case workflow
evals/
└── trigger-evals.json # Starter trigger tests for scope and regression checks
scripts/
├── build_release_artifacts.py # Sync plugin copy and build Claude/Codex ZIPs
└── validate_repo.py # Structural validation for paths, metadata, and eval coverage
Run the structural validator before publishing changes:
python3 scripts/validate_repo.pyGitHub Actions runs the same validator on pushes and pull requests.
If you change the skill description, trigger boundaries, or output format, update evals/trigger-evals.json in the same PR.
To build release artifacts:
python3 scripts/build_release_artifacts.pyThis creates:
dist/postmodernist-claude-skill-v<version>.zipfor Claude uploadsdist/postmodernist-codex-plugin-v<version>.zipfor Codex plugin distributiondist/SHA256SUMS.txtfor checksum verification
To publish a GitHub release automatically, bump plugins/postmodernist/.codex-plugin/plugin.json to the target version and push a matching tag such as v0.1.0.
Six theorists who saw the frame before it closed:
- Peter Barry — Beginning Theory
- Patricia Waugh — Metafiction
- Jean-François Lyotard — The Postmodern Condition
- Jean Baudrillard — Simulacra and Simulation
- Brian McHale — Postmodernist Fiction
- Roland Barthes — "The Death of the Author"
See CONTRIBUTING.md.
MIT — reality is open source. Debug responsibly.
This README was written with postmodernist.