Work Application

Work Application Skill

Job search automation with CV generation, scraping, analysis, and tracking. 4 HTML CV templates (classic, modern-sidebar, two-column, creative), job scraping across 5 French platforms (Free-Work, WTTJ, Apec, HelloWork, LeHibou), keyword-based scoring, deep page analysis, full multi-dimension report (skills/company/location/salary with market data), and candidature tracking via markdown table. Stdlib only (except Playwright for scraping/analysis). Supports local and Nextcloud storage.

Profiles: ~/.openclaw/data/work-application/ · Config: ~/.openclaw/config/work-application/config.json

Trigger phrases

Load this skill when the user says anything like:

• "generate my CV", "create a resume", "render my CV"
• "adapt my CV for this job offer", "tailor my resume"
• "scrape job offers", "find jobs", "search for devops positions"
• "analyze job offers", "rank these jobs", "score job matches"
• "deep analyze these jobs", "scrape and analyze job pages"
• "analyze this job offer", "generate a report for this URL", "report on this job"
• "track this application", "log candidature", "update application status"
• "list my applications", "show candidatures"
• "validate my CV", "check CV formatting"
• "show my profile", "what's in my master profile"

Quick Start

python3 scripts/work_application.py profile show
python3 scripts/work_application.py render --template classic --output cv.html
python3 scripts/work_application.py report "https://example.com/job-offer"
python3 scripts/work_application.py track list

Setup

python3 scripts/setup.py       # interactive: profile + permissions + scraper config
python3 scripts/init.py        # validate configuration

config.json - behavior restrictions (all destructive capabilities disabled by default):

Storage & data

Nextcloud storage (optional): when storage.backend = "nextcloud" is set in config, files are stored on the user's Nextcloud instance instead of locally. Authentication is delegated entirely to the openclaw-skill-nextcloud skill - this skill never handles Nextcloud credentials. The nextcloud skill must be installed separately.

Cleanup: python3 scripts/setup.py --cleanup

Uninstall: rm -rf ~/.openclaw/data/work-application/ ~/.openclaw/config/work-application/

Security model

Capability isolation

All capabilities are disabled or restricted by default. The agent cannot perform actions until explicitly enabled in config.json:

Credential isolation

This skill stores no secrets and requires no environment variables. config.json contains only behavioral flags and defaults - never credentials.

• Local storage (default): reads/writes files under ~/.openclaw/data/work-application/. No authentication needed.
• Nextcloud storage (optional): delegates all authentication to the openclaw-skill-nextcloud skill, which manages its own credentials (NC_URL, NC_USER, NC_APP_KEY in ~/.openclaw/secrets/nc_creds). This skill never reads, stores, or handles Nextcloud credentials directly - it imports NextcloudClient from the nextcloud skill at runtime. The nextcloud skill must be installed and configured separately (clawhub install nextcloud-files).
• Scraping (optional, allow_scrape=true): Playwright runs without authentication. Job pages are fetched as anonymous HTTP requests. Company review scraping (Glassdoor/Indeed) also runs unauthenticated - no login or API key is used.

Path traversal protection

All storage operations (local and Nextcloud) validate filenames through _validate_name():

• Rejects absolute paths (/etc/passwd, C:\...)
• Rejects traversal sequences (../, ..\\, bare ..)
• Rejects null bytes (\x00)
• Normalizes slashes and collapses redundant separators
• LocalStorage additionally resolves the final path and verifies it remains inside the storage root directory

HTML output safety

All profile fields are passed through html.escape() before HTML rendering. No raw user content is inserted into templates. This prevents XSS if the generated CV is served or shared.

Network boundaries

• No network calls by default - CV generation, analysis, tracking are fully offline
• Network calls only happen when allow_scrape=true, exclusively via Playwright (headless Chromium)
• Domains contacted when scraping: free-work.com, welcometothejungle.com, apec.fr, hellowork.com, lehibou.com (job platforms), plus the specific job offer URL provided by the user
• Domains contacted by report (optional, allow_scrape=true): glassdoor.fr, fr.indeed.com (company reviews - unauthenticated, anonymous)
• Nextcloud storage (optional): contacts the user's own Nextcloud instance via the nextcloud skill - no third-party server
• URLs built from user data (company names for review lookup) are properly URL-encoded
• No background network activity, no telemetry, no phone-home

File output safety

• Reports and CVs are written only to the configured storage directory (~/.openclaw/data/work-application/ or Nextcloud remote path)
• Filenames derived from user data (company names) are strictly sanitized to ASCII alphanumeric + hyphens
• Subdirectories (reports/, jobs/) are created inside the storage root only

Module usage

from scripts._profile import load_master_profile, load_adapted_profile, save_adapted_profile
from scripts._cv_renderer import render_cv
from scripts._validators import validate_profile
from scripts._tracker import log_application, list_applications, update_status
from scripts._report import generate_report, format_report_markdown, save_report

CLI reference

# Profile
python3 scripts/work_application.py profile show
python3 scripts/work_application.py profile validate

# CV Rendering
python3 scripts/work_application.py render --template classic --output cv.html
python3 scripts/work_application.py render --template modern-sidebar --color "#e63946" --lang en

# Job Scraping
python3 scripts/work_application.py scrape
python3 scripts/work_application.py scrape --platforms free-work,wttj

# Job Analysis
python3 scripts/work_application.py analyze

# Deep Analysis (scrape job pages for detailed matching)
python3 scripts/work_application.py deep-analyze
python3 scripts/work_application.py deep-analyze --max 10

# Report (full multi-dimension analysis of a single offer)
python3 scripts/work_application.py report "https://example.com/job-offer"

# Application Tracking
python3 scripts/work_application.py track list
python3 scripts/work_application.py track list --status en_attente
python3 scripts/work_application.py track add "Thales" "DevOps Engineer" --location Paris --salary "55-65k"
python3 scripts/work_application.py track update "Thales" entretien

# Config
python3 scripts/work_application.py config

Templates

Adapt CV for a job offer

from scripts._profile import load_master_profile, save_adapted_profile
from scripts._cv_renderer import render_cv
from scripts._validators import validate_profile

# 1. Load master profile
master = load_master_profile()
# 2. Agent adapts profile for the job (select relevant skills, rewrite summary, etc.)
adapted = adapt_for_job(master, job_description)  # agent logic
# 3. Validate
report = validate_profile(adapted)
if not report["valid"]:
    print("Errors:", report["errors"])
# 4. Save and render
save_adapted_profile(adapted)
html = render_cv(adapted, template="classic", color="#2563eb")

Scrape → Analyze → Track

# 1. Scrape jobs
from scripts._scraper import JobScraper, filter_jobs, deduplicate
scraper = JobScraper(config)
jobs = asyncio.run(scraper.scrape_all())
jobs = deduplicate(filter_jobs(jobs, config["scraper"]["filters"]))

# 2. Analyze
from scripts._analyzer import rank_jobs, select_top
ranked = rank_jobs(jobs, master_profile)
selected = select_top(ranked)

# 3. Track top matches
from scripts._tracker import log_application
for job in selected["cdi"][:5]:
    log_application(job["company"], job["title"], location=job.get("location",""))

Quick candidature update

from scripts._tracker import update_status, list_applications
update_status("Thales", "entretien")
active = list_applications(status="en_attente")

CV Templates

Status icons

Scraper platforms

Ideas

• Set allow_scrape: false + allow_export: true for a CV-only mode
• Use scraper with allow_tracking: true to auto-log the best matches
• Adapt CV per job offer, validate, render, then log the application
• Use readonly_mode: true for a safe demo mode

Notes

• Playwright dependency: Only needed for scraping. CV generation is stdlib only.
• Profile structure: Master profile contains ALL data. Adapted profile is a filtered subset for a specific job.
• Validators: Port of the JavaScript validators - same limits and thresholds.
• i18n: CV rendering supports French and English section headings.
• Print-ready: Generated HTML includes @media print rules and A4 page breaks.

Combine with

API reference

See references/api.md for CLI command details and profile schema.

Troubleshooting

See references/troubleshooting.md for common errors and fixes.