Set up
Skills are installed by placing their folder where your agent loads skills.
The download comes from github.com/solutionsDigibull/ISATVON-prompting itself, not from our server, so what you get is exactly what its author published.
Install instructions come from the README
README
synced Oct 26, 2026ISATVON Prompting applies the seven ISATVON elements as a prompting framework — like COSTAR, but the structure covers the answer as well as the ask. You write (or let the skill write) a prompt with seven sections; the model is required to return its response in the same seven-part structure, so you can see what it assumed, what it used, and whether it checked itself.
Raw prompt ──► I S A T V O N prompt ──► any AI platform ──► I S A T V O N response
(role/rules, sources, (task-as-understood, sources
method+self-check, used, verification, constraints
tools, constraints, honored, the deliverable,
output contract, meta) assumptions & confidence)
Works on ChatGPT, Claude, Gemini, Perplexity, Copilot, Grok, and anything else that reads markdown — it's just text.
Try it in your browser → — paste a raw prompt, get the seven-section version back, edit it, copy it.
Contents
- Quick example · ISATVON vs COSTAR · Installation
- The seven sections · When not to use it
- Learn: Your first prompt · Per-platform tuning · FAQ
- Reference: Per-element guide · COSTAR mapping · Response contract · Glossary · Which route to take
- Examples: all 7, by difficulty · index.json
- Build on it: API reference · Architecture
- Contribute: CONTRIBUTING · Writing an example · Governance · Security
Quick example
Before: Summarize this report for me, keep it short.
After: a seven-section prompt where "short" is 150 words with a declared fallback, the model must verify every sentence traces to the report before answering, and the reply opens with the task as understood — so a misread is visible in one line. Full conversion: examples/generic-summarize.md.
ISATVON vs COSTAR
| COSTAR | ISATVON | ISATVON adds |
|---|---|---|
| Context | S — Source | explicit "do not assume" boundary |
| Objective | I — Instructions | role + hard rules |
| Style / Tone / Audience | V — Variables | measurable limits + fallback |
| Response | O — Outcome | the reply itself in ISATVON structure |
| — | A — Automation | method + self-verification before answering |
| — | T — Tech stack | tool policy (search, code, no fabrication) |
| — | N — Notification | mandatory assumptions/confidence report |
Full mapping and when to use which: references/costar-comparison.md.
Installation
As an Agent Skill (Claude Code / Claude Desktop): see INSTALL.md. Then just ask: "convert this into an ISATVON prompt: …"
Zero-install: copy templates/prompt-template.md into any AI chat and fill in the placeholders. No agent required.
The seven sections
| Section | In your prompt | In the model's response |
|---|---|---|
| I — Instructions | Role, task, rules | The task as it understood it |
| S — Source | Context to use, gaps not to fill | Sources it actually used |
| A — Automation | Method + self-verification step | How it verified the result |
| T — Tech stack | Capabilities allowed/forbidden | — |
| V — Variables | Measurable constraints + fallback | Constraints honored or broken |
| O — Outcome | Exact response contract | The deliverable itself |
| N — Notification | Required meta-reporting | Assumptions, confidence, omissions |
Per-element guidance with good/weak examples: references/prompting-guide.md.
Repository layout
isatvon-prompting/
├── SKILL.md # Agent Skill: converts any raw prompt → ISATVON prompt
├── templates/
│ ├── prompt-template.md # copy-paste skeleton, works without any agent
│ ├── prompt-template-lite.md # I + O + N, for self-contained questions
│ └── response-template.md # the response contract on its own
├── references/
│ ├── prompting-guide.md # the 7 elements as prompt sections
│ ├── costar-comparison.md # COSTAR ↔ ISATVON mapping
│ ├── response-format.md # the structured-response contract
│ ├── domain-model.md # glossary: the terms this repo uses
│ └── flow-guide.md # convert or skip? full or lite? which section to fix?
├── examples/ # 7 before/after conversions + generated index
├── docs/ # architecture, API reference, FAQ, tutorials
├── web/ # the website and converter (every dependency lives here)
├── EVALS.md # recorded trigger-eval transcripts
├── skill-registry.json # generated manifest for agent discovery
├── INSTALL.md · CITATIONS.md · CHANGELOG.md · CONTRIBUTING.md · GOVERNANCE.md
├── RELEASING.md · SECURITY.md · CODE_OF_CONDUCT.md · CONTRIBUTORS.md · LICENSE
└── assets/banner.svg
The repo root has no package.json — copying a template into a chat needs no toolchain.
Tests, linters and the site all live in web/; see docs/architecture.md.
When not to use ISATVON prompting
- One-shot stylistic asks — a tweet rewrite or a tone change doesn't need seven sections; plain COSTAR (or nothing) is fine.
- Conversation — this structures deliverables, not chat.
- Agent loop contracts — if you need a bounded, machine-validated retry loop (JSON manifest, MCP tools, guaranteed exits), that's the ISATVON specification and its validator, not this repo.
Contributing
Example conversions are the most valuable contribution — see CONTRIBUTING.md and docs/examples-guide.md. Framework changes need evidence: GOVERNANCE.md.
Enterprise
ISATVON is an open framework by DigiBull. For enterprise prompt governance, analytics, and hosted orchestration, see DigiBull Enterprise.
Apache-2.0 · Citations & inspiration · Security · Changelog