Prompts Library
A catalog of battle-tested prompts (the "jef" library, after Jeffrey Emanuel, plus local additions). Each entry is a full prompt living in references/<name>.md.
How to use
- The user invokes
/jef-prompts <hint> (e.g. /jef-prompts find a deep bug, /jef-prompts make the UI world-class).
- Match the hint against the catalog below.
- Load the single best-matching
references/<name>.md and run it. If 2–3 are plausible, name them and ask which (or run the closest and mention the alternatives).
- If nothing matches well, say so and suggest the nearest options — don't force a poor fit.
Variant suffixes: -genius = multi-disciplinary first-principles depth; -alien = removes the governor on reasoning, paradigm-breaking angles (use sparingly, high-cost/high-variance).
Catalog
Debugging & correctness
| Prompt |
Use when |
bug-hunter |
Standard bug hunt after writing/changing code |
bug-hunter-genius |
Multi-disciplinary first-principles forensic debugging |
bug-hunter-alien |
Bugs beyond human cognitive categories (deep, exotic) |
system-weaknesses |
Analyze systemic weaknesses / failure modes |
deployment-verifier |
Verify a deployment is actually healthy |
e2e-pipeline-validator |
Validate an end-to-end pipeline works |
error-message-improver |
Make error messages clear and actionable |
cli-error-tolerance |
Harden a CLI against bad input / errors |
headless-webview-crumb-debugging |
Find a runtime hang in a Capacitor/WKWebView app with no inspector — env-gated crumb trail read via the a11y tree |
Performance
| Prompt |
Use when |
deep-performance-audit |
Thorough performance audit |
code-optimizer-alien |
Radical optimization (exotic algorithms, self-healing) |
Code quality, refactor & review
| Prompt |
Use when |
de-slopify |
Strip AI-slop / boilerplate, tighten code |
code-reorganizer |
Restructure / reorganize a codebase |
stub-eliminator |
Find and finish stubs / TODOs / placeholders |
peer-code-reviewer |
Peer-style code review |
dependency-audit |
Audit dependencies (risk, bloat, updates) |
test-coverage-gaps |
Find untested paths / coverage gaps |
api-contract-validator |
Validate API contracts / consistency |
Planning, ideation & synthesis
| Prompt |
Use when |
brainstorm |
Diverge-then-converge when unsure of approach |
idea-wizard |
Develop / pressure-test a single idea |
dueling-idea-wizards |
Two opposing perspectives debate an idea |
premortem-planner |
Premortem — surface how a plan fails before starting |
hundred-to-ten-filter |
Cut 100 options down to the best 10 |
plan-enhancer-alien |
Genius-level enhancements to an implementation plan |
project-opinion-elicitor |
Elicit strong opinions / direction on a project |
synthesize-info |
Synthesize scattered info into a coherent whole |
multi-model-synthesis |
Fan a question to multiple models, synthesize |
Onboarding & docs
| Prompt |
Use when |
deep-project-primer |
Get deeply oriented in an unfamiliar codebase |
readme-reviser |
Rewrite / improve a README |
UI
| Prompt |
Use when |
stripe-level-ui |
Push UI quality to Stripe-grade polish |
Git & workflow
| Prompt |
Use when |
git-committer |
Craft clean, well-structured commits |
| `land` |
Land work / close out a session (flywheel) |
agent-swarm-launcher |
Launch a multi-agent swarm on a task |
robot-mode-maker |
Turn a workflow into a deterministic "robot mode" |
Adding prompts
New prompts go in references/<name>.md and get a catalog row above. (The personal prompt-add skill automates dedup-checking and slug selection.)