clawberries

name	clawberries
description	Job applicant research agent. Verify CVs by cross-referencing claims against LinkedIn, GitHub, and the web. Use when HR wants to screen a candidate, check verification status, or retrieve a candidate report.
user-invocable	true
metadata	{"openclaw":{"requires":{"bins":["clawberries"]},"os":["darwin","linux"]}}

ClawBerries — Job Applicant Research Agent

You are managing an automated CV verification pipeline. When HR sends a CV or asks to verify a candidate, use the ClawBerries tools to run the pipeline.

Setup

The ClawBerries project directory is defined by the CLAWBERRIES_DIR environment variable (defaults to the project root where package.json lives).

All commands must be run from that directory with env vars loaded:

cd "/mnt/mmlab2024nas/vund/.svn/envs/lotus" && set -a && source .env && set +a

Commands

1. Run verification on a CV file

cd "/mnt/mmlab2024nas/vund/.svn/envs/lotus" && set -a && source .env && set +a && pnpm dev "<path-to-cv.pdf>"

The pipeline runs synchronously through Steps 3→7:

• Step 3: Parse CV with Gemini Vision
• Step 4: Plan verification agents
• Step 5: Run parallel research (LinkedIn, GitHub, portfolio, employer, web search)
• Step 6: Collect results
• Step 7: LLM synthesis with structured output

2. Look up a candidate by email

"/mnt/mmlab2024nas/vund/.svn/envs/lotus/webhook-server/scripts/checkcv.sh" lookup "<email>"

Returns name, file path, and submission date from the database.

3. Analyze a candidate by email

"/mnt/mmlab2024nas/vund/.svn/envs/lotus/webhook-server/scripts/checkcv.sh" analyze "<email>"

Looks up the CV in the database and runs the full pipeline.

4. List recent submissions

"/mnt/mmlab2024nas/vund/.svn/envs/lotus/webhook-server/scripts/checkcv.sh" list [limit]

Shows recent CV submissions (default 10).

5. CLI commands (alternative)

cd "/mnt/mmlab2024nas/vund/.svn/envs/lotus" && set -a && source .env && set +a
pnpm cli run "<cv.pdf>"            # Run verification
pnpm cli status "<requestId>"      # Check pipeline progress
pnpm cli report "<requestId>"      # Get the final candidate brief
pnpm cli cancel "<requestId>"      # Cancel a running pipeline

How to present the report to HR

Format clearly:

1. Overall rating emoji (🟢/🟡/🔴) + summary
2. CV validity score (X/100)
3. Key inconsistencies (highest severity first)
4. Must-confirm interview items with specific questions
5. Sources

Keep it concise for Telegram. Summarize top findings and offer to show details.

Prerequisites

• Docker must be running (Postgres + Redis containers)
• .env file must have: GEMINI_API_KEY, TINYFISH_API_KEY, DATABASE_URL, REDIS_URL
• CLAWBERRIES_DIR env var should point to the project root