/workana

Servy's front door to Gera's freelance deal-flow on Workana (a LATAM-first freelance marketplace — Spanish/Portuguese heavy, which is Gera's native-language edge). This skill finds fit-ranked opportunities and helps Gera win them, while keeping a hard read-only boundary: Servy surfaces and drafts, Gera submits.

This is the agency-arm intake for the Penguin Alley empire (references/penguin-alley-org-chart.md Arm 4). A Workana lead that converts to a real engagement graduates into /forge for the full build.

When to use

• Gera invokes /workana or any natural trigger ("find freelance work", "any gigs", "scan workana", "draft a bid for X").
• The daily routines/workana-scan.md Telegram digest surfaced something and Gera wants to act on it.
• Gera wants a targeted search ("find me RAG chatbot projects", "anything in Spanish about automation").

If Gera didn't ask and the routine didn't fire, don't manufacture work.

The connection (how it works)

Workana has no public API. The scanner uses the site's own data path: requesting a public /jobs browse URL with the header X-Requested-With: XMLHttpRequest content-negotiates to JSON instead of the HTML shell — exactly what a logged-out browser does when you scroll the public board. We read public listings, nothing more. Full mechanism + safety posture: references/workana.md.

Key tuning fact: the query= free-text search returns fresh, newest-first results; the skills= taxonomy filter returns stale ones. The scanner drives off query=. Config (queries, scoring weights, budget floor, language) lives in scripts/workana-scan.config.json.

Process

Step 1 — Scan

Run the scanner and surface the ranked digest.

• Default (new since last run): python3 scripts/workana-scan.py --new-only --limit 12
• Everything matching (peek, no state change): python3 scripts/workana-scan.py --all --limit 15
• Targeted: python3 scripts/workana-scan.py --query "rag chatbot" --query "n8n" --all
• Higher floor: python3 scripts/workana-scan.py --min-budget 500

Each line shows [score] title · budget · bids · payment✓ · country · recency (fit reasons) and a direct /job/<slug> link. Read the score signals for Gera:

• Few bids = better odds. A $1–3k project with 2 bids beats one with 90 bids.
• payment✓ = verified payment method = a real, more-serious client.
• Recency — "fresh"/"yesterday" projects are still open; "months ago" usually filled.
• Budget — fixed Over USD 3,000 and mid USD 1,000–3,000 are the sweet spot.

Recommend the 2–4 best by odds × fit × budget, not just raw score.

Step 2 — Deep-read a chosen project

When Gera picks one, fetch its full detail (the digest only carries a short description):

Run this exact command — the UA below is the scanner's own (UA in scripts/workana-scan.py); use it verbatim, do not improvise or truncate it:

curl -sL -A "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/124.0 Safari/537.36" \
  -H "X-Requested-With: XMLHttpRequest" -H "Accept: application/json" \
  "https://www.workana.com/job/<slug>"

or WebFetch the public https://www.workana.com/job/<slug> URL. Pull: the real requirements, scope, deliverables, client country/history, budget, and what would make a bid stand out.

Step 3 — Draft a tailored bid (Gera submits)

Write a short, personalized proposal in the project's language (Spanish or English — match it). Voice rails (references/voice.md + the LinkedIn-voice memory):

• Curious-learner register, warm and specific. No hype, no guru framing, no "I'm an expert in everything."
• Open by showing you understood their problem, not by listing your résumé.
• Anchor credibility in real Penguin Alley work where relevant: Cleia (full-stack AI app — Next.js + Supabase + Stripe + an AI assistant with per-workspace memory), agentic systems, RAG, automation pipelines, Claude/OpenAI integrations, WhatsApp/Telegram bots.
• Ask 1–2 sharp scoping questions that prove you've thought about the build.
• Keep it tight (a Workana proposal is a pitch, not an SOW). End with a clear next step.
• No Penguin Alley penguin imagery — this is a text proposal. PA branding is for PA-page social content only.

Output the draft for Gera to review and paste into Workana himself. Never auto-submit.

Step 4 — Track it (optional)

If Gera bids, append a line to agency/workana-pipeline.md (create if absent): date · project · budget · bid sent · status. This is the agency deal-flow ledger.

Step 5 — Convert

If a client replies and it becomes a real engagement, graduate to /forge (Phase 1 forge-discovery → spec → build → Codex review → deploy). For a formal SOW once scoped, hand the capture to pa-proposal-writer.

Boundary (non-negotiable)

• READ-ONLY on Workana. The scanner never logs in, never bids, never touches an account. Submitting a proposal is an outward action Gera approves and performs each time.
• Polite. The scanner uses realistic request volume + delays + a hard request cap. Don't hammer it; don't add a tight loop.
• SoftServe-separate. This is Penguin Alley / personal freelance income — 100% independent of Gera's SoftServe employment (references/penguin-alley-org-chart.md: the agency arm is clear, not in conflict). Never route SoftServe customer data or IP near a freelance bid.

Files

• Engine: scripts/workana-scan.py · Config: scripts/workana-scan.config.json · State: scripts/workana-seen.json (gitignored)
• Routine: routines/workana-scan.md + scripts/workana-scan-job.sh + launchd/com.servy.workana-scan.plist
• Wiki: references/workana.md · Agency context: references/penguin-alley-org-chart.md