Aptha AI

Spec toolkit

The Analyzer needs an Agent Functional Spec: a JSON description of what your agent is for, what it decides and which limits it must respect. Here is how to get one that the Analyzer will accept the first time: let your coding agent write it from your source, then check it.

Let your coding agent write it
Hand this prompt and the template to the coding agent you already use. It reads your agent's source and writes an Agent Functional Spec, then checks it with the same rules the analyzer applies, so the upload goes through without errors.
  1. 1Download the kit (one ZIP: prompt, template, schema, checker and a README) and unzip it next to your agent's repository.
  2. 2Paste the prompt from the kit into your coding agent (Claude Code, Cursor, Copilot, Codex…) and tell it where the repository is.
  3. 3It writes the spec and runs the checker until it prints READY FOR THE ANALYZER. Upload that spec with the source.
Download the kit (.zip)Individual files: Prompt Template Schema Checker

The checker needs Python 3.9+ and pip install jsonschema; it does not run your agent or use the network. It cannot tell whether a claim is true — the analyzer does that — but it removes every mechanical failure first.

Or fill in a template yourself
A blank template in either format, plus a guideline document covering every exact value the schema accepts and the cross-field rules that most often cause a rejected upload.
Agent Functional Spec template — WordFillable Functional Spec form. Fill it in, then convert it back to JSON below.Agent Functional Spec template — JSONThe same starting point as the raw Agent Functional Spec JSON, for editing directly against the schema.Authoring guidelines (PDF)Every exact enum value, the cross-field rules, and the mistakes that most often fail an upload.
Agent Functional Spec ⇄ Word converter
Render an Agent Functional Spec JSON document as a fillable Word form, or convert a filled-in Word document back to Agent Functional Spec JSON. Pure templating; no model is involved. Sign in to use the converter.
Agent Functional Spec → Word

Renders an Agent Functional Spec document as a human-readable, fillable form — pure templating, no model involved. Select an Agent Functional Spec JSON file below (or convert one from Word, on the right) to enable this.

Word → Agent Functional Spec

Upload a filled-out Functional Spec to convert it back to JSON — deterministic table-label mapping, no NLP. The result is used as your Agent Functional Spec document below automatically.