Parallel Enrichment

Bulk data enrichment via Parallel API. Adds web-sourced fields (CEO names, funding, contact info) to lists of companies, people, or products. Use for enriching CSV files or inline data.

Security Scan

The SKILL.md reads like a legitimate CLI guide for Parallel.ai enrichment, but the skill metadata omits required dependency and credential declarations (e.g., parallel-cli binary and an API key), creating an incoherence that warrants caution.

Purpose & Capability

The skill's stated purpose is to call the Parallel API via a CLI (parallel-cli) to enrich data. However, the registry metadata declares no required binaries and no credentials. A data-enrichment tool that talks to an external API normally requires: (a) the parallel-cli binary (or an install step) and (b) an API key or login. The absence of those declarations is inconsistent with the claimed purpose.

Instruction Scope

SKILL.md instructs the agent to read input CSVs, write output CSVs, preview rows (e.g., head -6), and to spawn sessions_spawn sub-agents to read and summarize output files. Those actions are within the expected scope for enrichment, but the guidance gives the agent the ability to read arbitrary file paths if misused—so confirm the agent's file-access boundaries and that spawned sessions only read expected files.

Install Mechanism

There is no install spec (instruction-only), which reduces direct install risk. However, the instructions assume the presence of 'parallel-cli' on PATH without declaring it as a required binary or providing install steps; this is an omission that affects reproducibility and security review.

Credentials

No environment variables or primary credential are declared. In practice, communicating with Parallel.ai typically requires authentication (API key or CLI login). The SKILL.md does not document how auth is provided (env var, config file, or interactive login), so required secrets could be hidden in platform config or left undocumented—this mismatch is important to resolve before use.

Persistence & Privilege

The skill does not request persistent installation, does not set always:true, and has no install-time hooks. Autonomous invocation is allowed (platform default). The main persistence concern is that instructions encourage writing files (target CSVs) and spawning sessions to read them—normal for the task but verify where outputs and any cached credentials are stored.

What to consider before installing

Before installing or using this skill: (1) Verify you have parallel-cli installed and know how it is authenticated—ask the skill author whether an API key (e.g., PARALLEL_API_KEY) or CLI login is required and where that credential is stored. (2) Treat the absence of declared binaries/credentials as a red flag: do not assume the skill is self-contained. (3) Run the CLI locally on a small non-sensitive dataset first to confirm behavior (what is sent to the network, monitoring URLs, and output format). (4) Be cautious about file paths: the skill instructs spawning sub-agents to read files—ensure those tools only access intended files and that no sensitive files are in the same directories. (5) If you need higher assurance, ask the publisher for provenance (author identity, source repo or install instructions) or prefer an officially published integration from Parallel.ai that documents auth and installation.

Bulk data enrichment that adds web-sourced fields to lists of companies, people, or products. Describe what you want in natural language.

When to Use

Trigger this skill when the user asks for:

• "enrich this list with...", "add CEO names to...", "find funding for these companies..."
• "look up contact info for...", "get LinkedIn profiles for..."
• Bulk data operations on CSV files or lists
• Adding web-sourced columns to existing datasets
• Lead enrichment, company research, product comparison

Quick Start

# Inline data
parallel-cli enrich run \
  --data '[{"company": "Google"}, {"company": "Microsoft"}]' \
  --intent "CEO name and founding year" \
  --target output.csv

# CSV file
parallel-cli enrich run \
  --source-type csv --source input.csv \
  --target output.csv \
  --intent "CEO name and founding year"

CLI Reference

Basic Usage

parallel-cli enrich run [options]

Note: There is no --json flag for enrich. Results are written to the target file.

Common Flags

Flag	Description
--data "<json>"	Inline JSON array of records
--source-type csv	Source file type
--source <path>	Input CSV file path
--target <path>	Output CSV file path
--source-columns "<json>"	Describe input columns
--enriched-columns "<json>"	Specify output columns
--intent "<description>"	Natural language description of what to find
--processor <tier>	Processing tier (see table below)

Processor Tiers

Processor	Use Case
lite-fast	Simple lookups
base-fast	Basic enrichment
core-fast	Standard enrichment
pro-fast	Deep enrichment (default)
ultra-fast	Complex multi-source enrichment

Examples

Inline data enrichment:

parallel-cli enrich run \
  --data '[{"company": "Stripe"}, {"company": "Square"}, {"company": "Adyen"}]' \
  --intent "CEO name, headquarters city, and latest funding round" \
  --target ./companies-enriched.csv

CSV file enrichment:

parallel-cli enrich run \
  --source-type csv \
  --source ./leads.csv \
  --target ./leads-enriched.csv \
  --source-columns '[{"name": "company_name", "description": "Company name"}]' \
  --intent "Find CEO name, company size, and LinkedIn company page URL"

With explicit output columns:

parallel-cli enrich run \
  --data '[{"name": "Sam Altman"}, {"name": "Satya Nadella"}]' \
  --source-columns '[{"name": "name", "description": "Person full name"}]' \
  --enriched-columns '[
    {"name": "current_company", "description": "Current company/employer"},
    {"name": "title", "description": "Current job title"},
    {"name": "twitter", "description": "Twitter/X handle"}
  ]' \
  --target ./people-enriched.csv

Using AI to suggest columns:

# First, get AI suggestions
parallel-cli enrich suggest \
  --source-type csv \
  --source ./companies.csv \
  --intent "competitor analysis data"

# Then run with suggested columns
parallel-cli enrich run \
  --source-type csv \
  --source ./companies.csv \
  --target ./companies-analysis.csv \
  --intent "competitor analysis: market position, key products, recent news"

Best-Practice Prompting

Intent Description

Write 1-2 sentences describing:

• What specific fields you want to add
• Context about the data (B2B companies, tech startups, etc.)
• Any constraints (recent data, specific sources)

Good:

--intent "Find CEO name, total funding raised, and number of employees for B2B SaaS companies"

Poor:

--intent "Find stuff about these companies"

Source Column Descriptions

When using --source-columns, provide context:

[
  {"name": "company", "description": "Company name, may include Inc/LLC suffix"},
  {"name": "website", "description": "Company website URL, may be partial"}
]

Response Format

The CLI outputs:

• A monitoring URL to track progress
• Status updates as rows are processed
• Final output written to target CSV

The target CSV contains:

• All original columns from the source
• New enriched columns as specified
• A _parallel_status column indicating success/failure per row

Output Handling

After enrichment completes:

1. Report the number of rows enriched
2. Preview the first few rows: head -6 output.csv
3. Share the full path to the output file
4. Note any rows that failed enrichment

Configuration File

For complex enrichments, use a YAML config:

# enrich-config.yaml
source:
  type: csv
  path: ./input.csv
  columns:
    - name: company_name
      description: "Company legal name"
    - name: website
      description: "Company website URL"

target:
  type: csv
  path: ./output.csv

enriched_columns:
  - name: ceo_name
    description: "Current CEO full name"
  - name: employee_count
    description: "Approximate number of employees"
  - name: funding_total
    description: "Total funding raised in USD"

processor: pro-fast

Then run:

parallel-cli enrich run enrich-config.yaml

Running Out of Context?

For large enrichments, save results and use sessions_spawn:

parallel-cli enrich run --source-type csv --source input.csv --target /tmp/enriched-<topic>.csv --intent "..."

Then spawn a sub-agent:

{
  "tool": "sessions_spawn",
  "task": "Read /tmp/enriched-<topic>.csv and summarize the results. Report row count, success rate, and preview first 5 rows.",
  "label": "enrich-summary"
}

Error Handling

Exit Code	Meaning
0	Success
1	Unexpected error (network, parse)
2	Invalid arguments
3	API error (non-2xx)

Common issues:

• Row failures: Check _parallel_status column in output
• Timeout: Use smaller batches or lower processor tier
• Rate limits: Add delays between large enrichments

Prerequisites

1. Get an API key at parallel.ai
2. Install the CLI:

curl -fsSL https://parallel.ai/install.sh | bash
export PARALLEL_API_KEY=your-key

References

• API Docs
• Enrichment API Reference