Upwork AI Agent Flowchart Proposal Appli

Upwork AI Agent (Flowchart + Proposal + Application Generator)

Before you run this

By default this skill uses local CSV files for its data (input and output) and writes local Markdown files for the proposal docs — nothing to connect, works offline, your data stays on your machine.

• Input/output CSVs: data/input.csv / data/output.csv
• Proposal docs: data/proposals/<title>.md — one file per job, rendered with the full template

If you'd rather read/write a Google Sheet instead, just say so and I'll switch you over. It's a one-time setup: you'll paste a Google service-account JSON (or authorize once), share your Sheet with that account's email, and give me the Sheet URL. Say "use Google Sheets" to start that, or "keep CSVs" (default) to just go.

If you'd rather have the proposal land in Google Docs (with a shareable link you can drop into the application), see Option C below.

What this does

Takes an Upwork job description and generates three assets in one shot:

1. Application copy — a short, punchy proposal body in the "I do this all the time, here's proof" style that high-converting freelancers use. Drops a link placeholder ($$$) where your proposal doc link goes, then swaps it in automatically once the doc is created (Google Docs mode) or left as-is (local mode).
2. Mermaid flowchart — a graph TD flowchart of how the system described in the job would actually be built. Paste it into any Mermaid renderer (or inline in a Notion doc, GitHub, etc.) to show the prospect you already thought it through.
3. Proposal doc — a full structured proposal (title, brief explanation, step-by-step build plan, left-to-right flow summary, about-me bullets). By default written as a local .md file to data/proposals/<title>.md. Opt in to Google Docs (see Option C) to get a shareable link instead.

The original workflow was an n8n chat agent with three sub-workflows wired as tools. This skill does the same thing — same prompts, same logic — but runs locally via Python scripts and stores results in CSV (or Sheets if you configure it).

When to trigger this

Trigger on any of:

• "generate upwork application"
• "write me a proposal for this job"
• "create flowchart + proposal for upwork"
• /upwork-ai-agent
• Pasting an Upwork job description and asking for application materials

Env vars to set before running

Required:

OPENAI_API_KEY=sk-...

Optional (local mode, no Google needed):

OPENAI_MODEL=gpt-4o-mini          # default, swap for gpt-4o if you want better output
ABOUT_ME="..."                    # override the default "about me" block used in proposals
SKILL_STORE=sheets                # switch from CSV to Google Sheets (see top of file)
SKILL_SHEET_URL=https://...       # required if SKILL_STORE=sheets

Required only to opt in to Google Docs output (both must be set — see Option C):

GOOGLE_APPLICATION_CREDENTIALS=/path/to/service-account.json   # must be type=service_account
PROPOSAL_TEMPLATE_DOC_ID=<doc-id-from-template-URL>

Optional Google Docs extras:

PROPOSAL_DRIVE_FOLDER_ID=...      # Google Drive folder to drop copied docs into
SKIP_GOOGLE_DOCS=1                # force local .md even when Google creds are set

Auto-detection: if GOOGLE_APPLICATION_CREDENTIALS and PROPOSAL_TEMPLATE_DOC_ID are both set, the skill uses Google Docs automatically. If either is missing it falls back to local .md. Set SKIP_GOOGLE_DOCS=1 to force local regardless.

Step-by-step procedure

Option A — single job (quick test or one-off)

Paste a job description and run:

cd /path/to/upwork-ai-agent-flowchart-proposal-appli
OPENAI_API_KEY=sk-... python scripts/run_agent.py --job "paste the full job description here"

Output prints to stdout: application copy, Mermaid code, and the path to the local .md proposal (or a Google Doc URL if Google creds are configured).

Option B — batch (process multiple jobs at once)

1. Open data/input.csv. It has two columns: job_title and job_description.
2. Add one row per job. Job description can be multiline — just wrap in quotes.
3. Run:

cd /path/to/upwork-ai-agent-flowchart-proposal-appli
OPENAI_API_KEY=sk-... python scripts/run_agent.py

4. Results land in data/output.csv with columns:
   • job_title
   • job_description
   • application_copy — the short proposal body ready to paste into Upwork
   • mermaid_code — paste into any Mermaid renderer
   • proposal_doc_url — absolute path to local .md file (or Google Doc URL if Google configured)
   • proposal_json — raw structured JSON from the proposal generator (for debugging)

Local .md proposals also land in data/proposals/<title>.md — one file per job.

Option C — Google Docs setup (opt-in, one-time)

If you want proposals as shareable Google Docs instead of local .md files:

1. Create a Google service account in Google Cloud Console, download the JSON key (type: service_account).
2. Enable Drive API and Docs API on the project.
3. Create a template Google Doc using {{titleOfSystem}}, {{briefExplanationOfSystem}}, etc. as placeholders (the same field names in the proposal JSON schema above).
4. Copy the PROPOSAL_TEMPLATE_DOC_ID from the template URL (the long string between /d/ and /edit).
5. Share the template doc + your target Drive folder with the service account's client_email.
6. Set GOOGLE_APPLICATION_CREDENTIALS and PROPOSAL_TEMPLATE_DOC_ID in your env and re-run. Both vars must be present for the Google path to activate.

Scripts reference

Node map (original n8n workflow → this skill)