---
description: Process project inputs into per-resource Markdown records and update the input index
argument-hint: <project-slug> [audio/text paths and/or inline text]
allowed-tools: ["Bash", "Read", "Write", "Edit", "MultiEdit", "Glob", "LS", "mcp__plugin_machine-business_free-whisper__*"]
---

Invoke the `process-input` skill to process raw audio/text inputs for one project slug.

The skill writes one resource record per input under:

```text
docs/inputs/<project-slug>/transcriptions/
```

and updates:

```text
docs/inputs/<project-slug>/index.json
```

A 3-phase pipeline:

1. **Classify & index check** — parse `<project-slug> [files...]`, classify each resource, and skip sources already `ready` in the project index.
2. **Transcribe** — transcribe only new audio resources with the free-whisper MCP. Text and inline inputs pass through unchanged.
3. **Write records & index** — write one `.md` per new ready resource, record failures in `index.json`, and stop.

If no files are passed after the project slug, use `docs/inputs/<project-slug>/raw/*`.

Hard rules: project slug required, source-path reuse, verbatim transcription, no fabricated content, no consolidated batch `.md`, no auto-chaining to other skills.

Final next-step rule: if the final response suggests the next slash command, it MUST print `/machine-business:business-proposal <project-slug>`.

Inputs to process:

$ARGUMENTS