aewing/skills · public

Curated agent skills.
Rigor with calm.

Eleven portable skills I actually use. Finish long-horizon work for real. Reason past plausible stories. Pressure-test products with honest synthetic panels. Write responses and interfaces people can scan. No project law, no product names, no private tooling — drop them into any agent.

Skills

Each skill is a folder with a SKILL.md — the portable Agent Skills format.

RIGOR CORE

execution

execution-rigor

Long-horizon execution that finishes for real: observable definition-of-done, per-task verification, shortcut tripwire scans, adversarial pre-completion audit.

read SKILL.md →
reasoning

truth-loop

Reasoning that converges on truth, not plausibility: hypotheses kept alive, evidence that can hurt, contradiction ledgers, a falsifier on every conclusion.

read SKILL.md →
language

goodtalk

Audit and recode instruction language so humans and agents stop accidentally pressuring each other. Research / standards / suggest / execute / recode / publish modes.

read SKILL.md →

MOVING & PROVING WORK

momentum

keep-going

Maintain progress on multi-step, long-running work: explicit remaining state, one causal unit at a time, milestone reviews, honest handoffs.

read SKILL.md →
focus

coherence-engine

Focus and pruning for long-horizon tasks: flow perception, cut what does not serve the core intent, autonomy with a clear reversible line.

read SKILL.md →
proof

verification-hygiene

Prove claims against the real artifact: the narrowest falsifying check, required consequences per change type, reject false green.

read SKILL.md →
review

adversarial-reviewer

Evidence-first review with one verdict — SHIP / BLOCKED / REWORK / MORE PROOF — highest blast radius first, smallest honest remediation.

read SKILL.md →
product review

focus-group

Run labeled synthetic user panels with causal participant selection, bounded complete transcripts, evidence-aware findings, and predictable project outcomes.

read SKILL.md →

RESPONSES & CRAFT

responses

wat

ADHD-friendly response overlay: answer first, one screen, evidence kept, one next action. For status, summaries, and handoffs.

read SKILL.md →
fun

roast-review

Evidence-backed critique: funny, ruthless, specific reduction to the target's weaknesses, plus a concrete escape plan.

read SKILL.md →
craft

tui-design

Terminal UI craft: layout paradigms, keyboard and focus models, semantic color, animation, anti-pattern and compatibility checklists.

read SKILL.md →

Install

Pick the path that matches your agent.

CLAUDE CODE — PLUGIN MARKETPLACE1-2-3
# in Claude Code
/plugin marketplace add aewing/skills
/plugin install rigor-skills@aewing-skills
ANY AGENT — COPY INSTALLERALT
git clone https://github.com/aewing/skills.git
cd skills && ./install.sh

# options
./install.sh --list
./install.sh --dry-run
./install.sh --target ~/.config/your-agent/skills

Marketplace users update with /plugin marketplace update aewing-skills. Install-script users git pull and re-run ./install.sh when a skill changes. The repo self-governs: its language standards live in .goodtalk/standards.json.