Senso Ingest for ClawMax

This skill wraps the upstream senso-ingest workflow with ClawMax conventions.

Use it when the agent should add evidence into Senso in a way that is easy for other ClawMax agents to find and reuse.

Use This Skill For

• screenshots for bug triage
• PDFs and notes for research
• support evidence like transcripts and user summaries
• store or operations evidence for retail workflows
• raw markdown brief creation in Senso

Core Rule

Always use the upstream Senso CLI behavior, but standardize the structure and handoff so downstream agents know what to search for.

Prerequisites

• SENSO_API_KEY is set
• senso CLI is installed
• always use --output json --quiet

ClawMax Naming Pattern

When creating or uploading content, use a title pattern like:

<template> | <workflow> | <artifact-type> | <short-subject>

Examples:

• visual-qa-lab | visual-intake | screenshot | checkout-modal-overlap
• research-desk | evidence-intake | pdf | hackathon-brief
• customer-signal-desk | signal-ingest | note | call-summary-042

Preferred Folder Pattern

If a folder is available or can be created, prefer:

ClawMax / <template> / <workflow>

Examples:

• ClawMax / Visual QA Lab / visual-intake
• ClawMax / Real-Time Research Desk / evidence-intake

If folder creation is too expensive for the current task, keep the title pattern strict and continue.

Ingestion Workflow

1. Determine artifact type

Choose one:

• file upload
• raw text / markdown creation
• replace existing content

2. Determine target structure

Before ingesting, decide:

• template name
• workflow name
• artifact type
• short subject
• target folder if needed

3. Ingest

Use the upstream commands:

senso ingest upload <file1> ... --output json --quiet

or:

senso kb upload <file1> ... --folder-id <kb_node_id> --output json --quiet

or raw text:

senso kb create-raw --data '{...}' --output json --quiet

4. Verify processing

Always verify after ingest:

senso kb sync-status --output json --quiet

If a specific content item was created:

senso content get <content_id> --output json --quiet

5. Confirm searchability

Run one search that should find the new content:

senso search "<relevant query>" --max-results 1 --output json --quiet

Required Handoff Output

After successful ingest, report back in this structure:

1. what was ingested
2. where it was placed
3. content IDs created or reused
4. whether processing is complete
5. what query downstream agents should use to find it

Example:

Ingested:
- 2 screenshots for Visual QA Lab / visual-intake

Placement:
- Folder: ClawMax / Visual QA Lab / visual-intake

Content IDs:
- cnt_123
- cnt_456

Status:
- processing complete

Suggested downstream query:
- "checkout modal overlap mobile screenshot"

Anti-Patterns

• do not upload without a consistent title/subject pattern
• do not stop at upload_pending if the next agent depends on retrieval
• do not return vague “uploaded successfully” with no IDs or search hints
• do not invent new folder patterns per template when reuse is possible