The filesystem is the library. The app is the search layer.

Perfect Prompts is a searchable library of prompts, rules, standards, and agent assets that lives in the filesystem - not in a proprietary database. Browse it normally in Git or your editor, or use the desktop app to search in plain English and take action on the best matches in one click.

Your personal filesystem-first library for prompt engineering, context engineering, agent control, reusable AI workflows, research methodology, and LLM tooling.

Two easy ways to use it:

1. Open the repository normally in your file manager, editor, terminal, or GitHub and browse/copy the artifacts directly.
2. Launch the Perfect Prompts desktop application when you want indexed search, batch queries, previews, or GUI-based add/remove operations.

The GUI does not own the library. The files do.

Current release: v2.0.2

Repository organization

The repository is now organized by what an artifact is, with format/runtime differences below that level instead of dominating the root structure.

Perfect_Prompts/
├── Agent_Instructions/      persistent agent operating instructions
├── Guidelines/              cross-cutting guidance
├── Methodologies/           reusable interaction/design methods
├── Personas/                role/persona definitions and target formats
├── Prompts/
│   ├── Templates/           standalone and specialized prompt templates
│   ├── Portable/            plaintext, structured-text, and JSON forms
│   └── Runtime_Bindings/    Python, TypeScript, Rust, and Go implementations
├── Rules/                   scoped technical/scientific/domain rulesets
├── Skills/                  reusable procedural skill packages
├── Standards/
│   ├── APEX/
│   ├── Architecture/
│   ├── NASA/
│   └── Research/
├── External_References/     separately maintained/reference prompt libraries
├── Application/             optional Perfect Prompts desktop application
├── REPOSITORY_MAP.md        old → new path map and organization rationale
├── install.py               optional desktop-app installer
├── LICENSE
└── README.md

See REPOSITORY_MAP.md for the migration map from the older organically grown layout.

Browse it directly

No application is required.

If you want an ordinary prompt you can start in Prompts/. If you need a persistent agent instruction, start in Agent_Instructions/. For procedural capability packages use Skills/; for scoped reasoning constraints use Rules/; for architecture/research/engineering authority use Standards/; and for role specialization use Personas/.

The representation split inside Prompts/ is deliberate:

• Portable/ is for directly inspectable/copyable prompt and context forms.
• Runtime_Bindings/ keeps prompt implementations embedded in Python, TypeScript, Rust, and Go so application code can use them without translating them into a detached format.
• Templates/ contains standalone/specialized prompt-template families.

Parallel implementations are not automatically assumed to be identical. They remain inspectable as their own artifacts.

Desktop application

The desktop application is also called Perfect Prompts and uses the existing Perfect Prompts logo as its native application icon.

Its intentionally small workflow is:

Search → inspect/open/copy → use the artifact
                   ↘ remove when needed

Batch → run several independent searches

Library → browse the real filesystem → add/remove/open artifacts

The primary feature is Prompt Beacon, a Perfect Prompts-specific version of the Beacon indexing/query system. It provides:

• local SQLite FTS5 indexing;
• broad prefix search for ordinary terms;
• strict quoted-phrase search;
• BM25 ranking;
• path, type, area, runtime, and source filters;
• single-query JSON export;
• independent batch-query execution and exports;
• extraction from ordinary text/source files, notebooks, Office/OpenDocument files, ZIP/skill packages, and PDFs when PDF support is installed;
• incremental synchronization after files are added, changed, moved, or removed.

The search database lives under .perfect-prompts/ and is disposable. Delete it at any time; the application rebuilds it from the repository.

Filesystem and GUI stay in sync

You can manage the library either way:

• add, move, edit, or remove files using your OS file manager, Git, terminal, or editor;
• add or remove artifacts from the Library tab in the GUI;
• remove a search result directly from the Search workspace;
• press Sync at any time, or let the application synchronize automatically while it is running.

The GUI uses the same repository paths. There is no private application-only artifact store.

Install the desktop application

From the repository root:

python install.py

The installer creates an isolated environment under Application/.venv/, installs the GUI dependencies, builds the initial search index, and creates native launchers where supported:

• Linux application menu entry;
• Linux desktop launcher;
• Windows Start Menu shortcut;
• Windows desktop shortcut.

The launcher uses the supplied Perfect Prompts icon.

After installation you can also run the programmatic CLI:

perfect-prompts-cli sync
perfect-prompts-cli query "session handoff" --source-scope project
perfect-prompts-cli batch --queries 'architecture, "prompt template", context synthesis'
perfect-prompts-cli add ./new-skill --to Skills
perfect-prompts-cli remove Prompts/Portable/Plaintext/old-prompt.md

What is in the library

The corpus includes multiple generations of material developed through active use, including:

• agent and multi-agent prompts;
• project scaffolding and task-generation prompts;
• context builders and session-state systems;
• rewriting, citation, mathematical-formatting, Mermaid, reasoning, critique, and analysis prompts;
• persistent AGENTS-style instructions;
• Guided Autonomy and emergence-oriented methodologies;
• software, research, architecture, documentation, NASA-derived, and APEX standards;
• scientific and technical rulesets;
• personas in several target formats;
• reusable skill packages;
• Python, TypeScript, Rust, Go, plaintext, JSON, and structured representations;
• scripts and examples that accompany those artifacts;
• external/reference prompt and skill libraries.

This remains a working corpus. The reorganization improves navigation without pretending every historical artifact is current or every parallel implementation is equivalent.

Design rule

Meaning first, representation second; filesystem first, application second.

A person should be able to understand and use the repository by browsing it normally. The desktop application exists to make a large library faster to search and manage, not to make the files dependent on the application.