tecla-offers

name	tecla-offers
description	Tecla (app.tecla.io) as a source node in GARY's application network — a LATAM→US remote tech-talent marketplace (USD-paid) where applying is profile-based + AI-matched behind a 4-stage vetting/English gate. How to source/triage REMOTE-worldwide roles and why apply is NOT auto-fillable (open-and-hand-off, the user's click). Use when sourcing or deciding whether to apply to Tecla offers.

Tecla — source node

Candidate-agnostic. Canonical rules: docs/operating-rules.md; per-board runbook: docs/career-ops-map.md §4.

Tecla (app.tecla.io / tecla.io) is one website node in GARY's application network. The canonical candidate rules, hard rules (two-way-fit gate), the per-offer workflow, and CDP/network architecture live in gary-job-search + docs/operating-rules.md; every candidate fact (stack, YOE, gap skills, comp floor, English level, EEO, standard answers) is DATA — query the NotebookLM RAG (via notebooklm-ai-plugin) or read config/profile.yml / cv-data.md as the derived cache. This skill does NOT duplicate them — only what is Tecla-specific.

What Tecla is

A LATAM→US remote/nearshore tech-talent marketplace ("Hire the best tech & AI talent across the US and LATAM"). It connects vetted Latin-American engineers (50k+ across 18+ countries) to remote roles, mostly at US companies, paid in USD. Candidate-free (no fees to join or get hired). Three lines: nearshore staff augmentation, AI recruiting, executive search. It is AI-match driven ("create a profile and let opportunities come to you"), not an open free-apply ATS board. This is a vetting-gated marketplace: the candidate is selected into a shortlist, not free-applying to a form.

(1) BLOCKS / anti-bot

• Login-gated, no anonymous API. The job board lives at app.tecla.io/jobs, a JS-rendered SPA; the public HTML exposes no anonymous job feed and probed API paths (/api/jobs, etc.) return empty/404. There is NO anonymous REST API to pull offers from (unlike getonbrd/himalayas). To see jobs you must be a signed-in candidate with a completed profile.
• Vetting gate (4 stages). The pipeline is gated by AI screening → technical assessment → recruiter interview (tests English fluency) → reference checks. The recruiter/English screen is an unavoidable human step and is higher-risk for a candidate below the English level a posting demands — flag any posting requiring above the candidate's DATA level (config/profile.yml → English level).
• One-time completed-profile prerequisite. Tecla yields nothing until the user has a completed candidate profile (signup + CV + profile fields) — completeness drives AI matches. Surface this prerequisite ONCE, then treat it as done.
• Mitigation: since sourcing is login-gated, source over the logged-in automation browser (debug Chrome, default port 9333 — DATA in config/profile.yml → job_search.automation_browser), NOT anonymous WebFetch. The user's personal browser is OFF-LIMITS. Serialize with the other CDP nodes.

(2) AUTOMATION — profile-based one-click apply, vetting-gated → NOT auto-fillable (open-and-hand-off)

Classification: one-click profile apply behind a multi-stage vetting + English gate (same family as getonbrd "Postular" / VanHack — NOT an anonymous ATS form). Concretely:

1. Apply = profile-based one-click / expression of interest — it submits the saved Tecla profile + CV, no per-role ATS form fields to auto-fill → nothing for the filler engine to do (0% auto-fill).
2. 4-stage vetting shapes the outcome (AI screen → technical assessment → recruiter/English interview → reference checks) — the recruiter/English screen is an unavoidable human step.
3. That click + the screen behind it = the application → per the hard rules it is the user's click, never the agent's. Recruiters/AI forward shortlists to companies, so there is no anonymous-ATS surface to partially fill (unlike LinkedIn external-ATS / JazzHR).

• Auto-fill feasibility: NONE. Recommended pattern: open-and-hand-off (manual). GARY's only automatable contribution is finding/triaging matches; profile setup, the apply click, the English/recruiter screen, and assessments are all the user's. The agent (job-apply-prep) OPENS each match in the automation browser and STOPS.

(3) PAGINATION (inside the logged-in job list)

• Jobs surface two ways: (a) AI match push — the more complete the profile, the more matches; and (b) a filterable job list (pull) by role, seniority, and work arrangement (remote). A "Salary Wizard" exposes USD comp data.
• No documented URL-param search contract to drive headlessly. Treat /jobs as a logged-in UI surface, not a queryable endpoint. Filter in-app for remote + senior + the candidate's stack, paginate the whole logged-in list, and apply the ≤1-week recency rule yourself from the card date (Tecla has no reliable public date filter).

(4) WEBFETCH FILTERING

• Tecla is login-gated → anonymous WebFetch cannot read the job feed or a real JD (SPA + no public API). Sourcing is logged-in CDP over the automation browser, NOT WebFetch. Do not rely on a guest WebFetch to validate an offer — open the real JD inside the signed-in session and validate location, comp, YOE, English requirement, and open/closed status there.

Triage — two-way-fit gate, REMOTE-WORLDWIDE scope

Run the standard gate (docs/operating-rules.md §1 hard rules), but do NOT geo-limit by the company's country. Tecla's whole purpose is connecting LATAM talent to remote roles (often US companies), so the company being US/elsewhere is irrelevant — only "remote-workable from the candidate's country" matters. Apply to ANY-country remote role meeting the mandatories:

• Remote (Tecla roles generally are) AND hireable remotely from the candidate's country. Reject ONLY region-LOCKED roles that explicitly exclude the candidate's country (e.g. "US-only / E-Verify / US work authorization", "EU-residents only"). A US company hiring remote-LATAM/ Americas/worldwide = PASS.
• Comp: ≥ the comp floor in config/profile.yml (Tecla pays USD — use the Salary Wizard / posting band; convert annual → monthly). Unknown ≠ below floor (keep, confirm early).
• Seniority / profile fit: Tecla skews senior (good fit). Reject roles above the candidate's total YOE and backend-specialist roles demanding more backend years than the candidate has (total YOE ≠ domain YOE) — read the profile from the NotebookLM RAG / cv-data.md.
• Req-fit against the candidate's real stack (RAG / cv-data.md). Reject roles whose CORE is a gap skill (gap skills are DATA in the RAG / cv-data.md).
• English: the recruiter interview tests spoken English — flag any posting demanding a level above the candidate's DATA level (config/profile.yml). Company already in data/applications.md → skip (one app per company).

How the orchestrator uses this node

1. Lower-priority, CDP-serialized + setup-dependent node. No public API + login-gated + zero auto-fill → it sits behind the auto-fillable channels (LinkedIn Easy Apply / external ATS) that carry the auto-apply target; Tecla contributes 0% auto-fill (open-and-hand-off only).
2. One-time setup reminder (surface once): Tecla yields nothing until the user has a completed candidate profile (signup + CV + profile fields). Prompt them to finish it before this node is worth running.
3. Sourcing (logged-in CDP, serialized): filter /jobs to remote + senior + the candidate's stack, triage with the gate above (no geo-limit), and merge survivors into the canonical data/offers-master.md (per operating-rules.md §3): Company | Title | Remote-eligible? (PASS unless region-locked-excl-candidate) | Comp USD/mo | Apply URL | fit note.
4. Hand survivors to job-apply-prep, which OPENS each match in the automation browser and STOPS (the user's profile-apply click + any English/recruiter screen). After the user confirms an apply, flip the tracker row Evaluated → Applied (never before).

Cross-references (don't duplicate here)

• Candidate facts, skills+YOE, gap skills, comp floor, English level, EEO, standard answers → NotebookLM RAG (notebooklm-ai-plugin) / config/profile.yml / cv-data.md.
• Hard rules, per-offer workflow, network architecture, CDP serialization → gary-job-search + docs/operating-rules.md; per-board runbook → docs/career-ops-map.md §4.7.
• Sibling profile-apply / vetting-gated nodes (one-click, no fields, hand-off) → vanhack-offers, xpertdirect-offers, getonbrd-offers.