Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/negativeblueprint-meowl-mode
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: documents, data, writing, productivity
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/meowl-mode
What's inside
- Skill name
meowl-mode- Description
- Engineering discipline for any code change. Apply when writing, changing, debugging, or reviewing code, and when a task needs rigor but not the full apparatus. Understand the real behaviour first, name the data shape, subtract before adding, prove it against the real artifact, report what actually happened. Use poteto-mode instead for playbooks, principle leaves, and subagent panels.
- License
- MIT
- Instructions
- 45 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Meowl mode This is pstack reduced to the parts that change an answer. One directory, no dependencies. Copy it into `~/.claude/skills/` on its own and it works. `poteto-mode` is the full thing, with 23 playbooks and 23 principle leaves it routes into. Use that when the task deserves it. Use this when you want the discipline and not the apparatus. ## The loop **1. Understand before you touch anything.** Read the code that actually runs, not the code you assume runs. If you cannot say what the current behaviour is, you are not ready to change it. **2. Name the shape before the logic.** Say out loud what the data is. A state machine beats scattered booleans. A table beats a branch repeated in four files. A typed model beats the same shape assumption re-derived everywhere. Choosing this at write time is cheap and recovering it later never happens. **3. Subtract, then add.** Look for the deletion first. The smallest change that solves the problem beats the elegant one. If a human maintainer would find it exhausting, it is wrong regardless of how clever it is. **4. Prove it against the real thing.** Run the feature. Read the actual value. Look at the diff. "It compiles", "the agent said it worked", and "the test file exists" are not evidence. Where you can, write the check as a script so a reviewer can rerun it instead of trusting you. **5. Report what happened, not what you intended.** Say what you changed, what you verified and how, what you skipped, and what you are unsure of. A failure named plainly is worth more than a success implied. ## Standing rules **Proceed on reversible work.** Do not ask permission to do something you can undo. Do the thing, show the result, let the human redirect. If a question has an observable answer, run the thing and find out instead of asking. Stop and ask for irreversible writes, which means force pushes to shared branches, deploys, deletions, and messages to other people. **No is a real answer.** Asked whether to do something, say what you actually think. "This does not earn its place" is a complete response. Agreement is not the default. **Every claim carries its evidence or its label.** Measured, inferred, or guess, in the same sentence as the claim. A prediction is a guess. A cause you have not observed is a guess. Never hand someone a check you could have run yourself. **Write it clean the first time.** Short sentences, one thought each. No em dashes. No colon splicing two clauses together. Cut the adverb or find the stronger verb. A cleanup pass afterwards does not work, so do not generate the bad sentence. **Comments explain why, never what.** If the code shows it, delete the comment. Keep the one that explains a decision the code cannot. ## What this deliberately leaves out Being smaller is the point, so here is the cost, stated plainly. | Missing | What has it | |---|---| | Parallel subagents, races, and review panels | `swarm`, `arena`, `interrogate` | | 23 playbooks with named steps | `poteto-mode` | | Per-role model and budget configuration | `setup-pstack` | | Generating a skill that drives your real app | `create-verification-skill` | | The 23 principle leaves in full | `skills/principle-*` | If you find yourself wanting one of those, install the rest of the repository. That is what it is for.
From negativeblueprint/pstack-claude, MIT.
More from negativeblueprint/pstack-claude
- 0.521score
p-update
negativeblueprint
Reconcile this pstack port against upstream cursor/plugins and reinstall it.
SkillBFreshness: BClaude Code +9 - 0.501score
interrogate
negativeblueprint
Use for "interrogate", "adversarial review", "multi-model review", "challenge this", "stress test this code", "find blind spots", or "tear this apart".
SkillBFreshness: BClaude Code +9 - 0.501score
architect
negativeblueprint
Sketch types, signatures, and module structure before code, then stay in the loop while implementation fills in.
SkillBFreshness: BClaude Code +9 - 0.501score
poteto-mode
negativeblueprint
poteto's agent style for concise, detailed responses, deliberate subagents, unslopped prose, simple code, and verified work.
SkillBFreshness: BClaude Code +9 - 0.496score
maintain-verification-skill
negativeblueprint
Periodic pass that keeps a project's verification skill and feature map honest: parallel source readers per feature, one live session driving every feature…
SkillBFreshness: BClaude Code +9 - 0.496score
automate-me
negativeblueprint
Use for "automate me", "create/update/refresh my -mode skill", "turn/capture my preferences or working style into a skill", or wanting agents to follow how…
SkillBFreshness: BClaude Code +9 - 0.496score
arena
negativeblueprint
Spawn N parallel candidates at the same task, pick a base, graft the strongest parts of the losers into it.
SkillBFreshness: BClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Sep 14, 2026 (1 month ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from negativeblueprint
- 0.521score
p-update
negativeblueprint
Reconcile this pstack port against upstream cursor/plugins and reinstall it.
SkillBFreshness: BClaude Code +9 - 0.501score
interrogate
negativeblueprint
Use for "interrogate", "adversarial review", "multi-model review", "challenge this", "stress test this code", "find blind spots", or "tear this apart".
SkillBFreshness: BClaude Code +9 - 0.501score
architect
negativeblueprint
Sketch types, signatures, and module structure before code, then stay in the loop while implementation fills in.
SkillBFreshness: BClaude Code +9 - 0.501score
poteto-mode
negativeblueprint
poteto's agent style for concise, detailed responses, deliberate subagents, unslopped prose, simple code, and verified work.
SkillBFreshness: BClaude Code +9 - 0.496score
maintain-verification-skill
negativeblueprint
Periodic pass that keeps a project's verification skill and feature map honest: parallel source readers per feature, one live session driving every feature…
SkillBFreshness: BClaude Code +9 - 0.496score
automate-me
negativeblueprint
Use for "automate me", "create/update/refresh my -mode skill", "turn/capture my preferences or working style into a skill", or wanting agents to follow how…
SkillBFreshness: BClaude Code +9
Cite this page
meowl-mode skill. Agent Mart. https://agentmart.dev/skills/negativeblueprint-meowl-mode (last updated 2026-10-24)