construction-agent
Skill121 starsChanged 5 months agoMITView on GitHub
Construction company AI agent — orchestrates email (Google Workspace), project management (JobTread Pave API), and file routing for a construction business. Handles email triage, attachment routing to JobTread, project status, task management, and daily briefings via Telegram. Reference implementation for a construction/trades business with an existing email→Drive→PM pipeline.
- Reads credentials
In this repository15 other files it gives its agents.
- agent-cheatsheet-builderSkill
- crm-liteSkill
- estimator-engineSkill
- financial-analysisSkill
- grasshopper-voicemail-monitorSkill
- hermes-production-opsSkill
- lead-to-paymentSkill
- mailchimp-integrationSkill
- marketing-campaign-builderSkill
- outlook-graphSkill
- quickbooks-onlineSkill
- small-business-ai-transformationSkill
- small-business-website-specSkill
- stripe-link-cliSkill
- stripe-paymentsSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 3 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 41
Every action follows the same philosophy as all Mom-n-Pop skills:
- Very specific: line 109
All set in ~/.hermes/.env (chmod 600):
- Distinctive: line 145
If you use Procore, Buildertrend, CoConstruct, or another construction PM system instead of JobTread:
- Very specific: line 148
Update the query/mutation structure in scripts/drive_to_jobtread.py to use your PM system's file upload endpoint
- Very specific: line 149
Update scripts/daily_briefing.py to query your PM system's task/project schema
- Distinctive: line 150
Update the environment variables to match your PM system's auth method
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Construction Agent — Email → Project Management Pipeline
Overview
A reference implementation of an AI agent that runs the back-office of a construction company. The agent orchestrates between three systems a construction business already uses:
- Email (Google Workspace / Gmail) — where leads, estimates, and project documents arrive
- File storage (Google Drive) — where email attachments are auto-dumped by an existing script
- Project management (JobTread, or similar PM system) — where files and tasks need to end up
The agent bridges the gap: it monitors Drive for new files, downloads them, uploads them to the PM system, and notifies the owner — all through text messages the owner already reads.
For who: A construction or trades business (general contractor, remodeler, specialty trade) where:
- The owner gets project documents via email
- Email attachments are dumped to a Drive folder (by a separate script or filter)
- Files need to get into a PM system (JobTread, Procore, Buildertrend, etc.)
- The owner doesn't want to manually download and re-upload everything
When to Use
- A customer or subcontractor emails a project document (plans, permits, invoices, photos)
- The owner asks "what's in the inbox?" or "what came in today?"
- A scheduled cron checks for new files to route to the PM system
- The owner wants a morning briefing: today's tasks, deadlines, new emails
- The owner needs a project status check from the PM system
The Pattern
Every action follows the same philosophy as all Mom-n-Pop skills:
- Agent drafts — the agent reads email, prepares file uploads, drafts replies
- Owner approves — the owner gets a Telegram message, confirms before anything is sent
- Agent executes — only after approval does anything go out
The exception is file routing (Drive → PM system), which runs autonomously on a cron because it's non-destructive (files are moved, not deleted; originals are preserved). The owner still gets a notification of what was routed.
Core Workflows
1. Email → Drive → Project Management Pipeline
This is the main pipeline. Many construction businesses already have a script that dumps email attachments to a Drive "raw inbox" folder. The agent completes the circuit:
Email arrives → existing script dumps attachment to Drive →
Agent polls Drive folder (every 15 min via cron) →
Downloads attachment locally (temp) →
Uploads to PM system via API →
Deletes local temp file →
Moves Drive file to "processed" folder →
Sends Telegram notification to ownerWhy the agent does this: The previous setup got files into Drive, but getting them from Drive into the PM system required manual download + upload. The agent automates that last mile.
2. Email Triage
- Reads Gmail inbox (via Google API)
- Summarizes unread emails
- Drafts replies (dual-approval: agent drafts, owner approves via Telegram or Gmail)
- Flags urgent items
3. Project Management Queries
- Query projects, tasks, schedules from the PM system
- Create/update tasks
- Daily briefing: today's tasks, upcoming deadlines
- Weekly summary: project progress, completed tasks
4. Daily Briefing (cron, 8 AM)
- New emails from overnight
- Today's tasks from the PM system
- Upcoming deadlines
- Files processed
API Integrations
Google Workspace
- Auth: OAuth2 (one-time browser flow, token persists at
~/.hermes/cache/google_token.json) - APIs: Gmail, Google Drive, Google Calendar
- Scopes:
gmail.readonly,gmail.send,drive.readonly,drive.file,calendar.readonly - Setup: See
references/google-workspace-setup.md
JobTread (Pave API)
JobTread is a construction-specific PM system. The Pave API pattern applies to similar construction PM tools (Procore, Buildertrend, etc.) — adapt the query language to your specific system.
- Endpoint:
POST https://api.jobtread.com/pave - Auth: Grant key (from JobTread Settings → Integrations → API)
- Query Language: Pave (similar to GraphQL, but not identical)
- Grant keys expire after 3 months of inactivity — must use regularly
- Setup: See
references/jobtread-api-setup.md
Environment Variables
All set in ~/.hermes/.env (chmod 600):
# Google Workspace
GOOGLE_CLIENT_ID=your_client_id_here
GOOGLE_CLIENT_SECRET=your_client_secret_here
GOOGLE_MAILBOX=info@yourdomain.com
# JobTread (or your PM system)
JOBTREAD_GRANT_KEY=your_grant_key_here
JOBTREAD_ORG_ID=your_org_id_here
# Google Drive (for attachment pipeline)
GOOGLE_DRIVE_RAW_FOLDER_ID=your_raw_folder_id
GOOGLE_DRIVE_PROCESSED_FOLDER_ID=your_processed_folder_idCron Jobs
| Name | Schedule | What |
| Daily Briefing | 0 8 * * * | Email summary, today's tasks, deadlines |
| Drive Poll | */15 * * * * | Check Drive for new attachments → upload to PM system |
| Weekly Summary | 0 9 * * 1 | Project progress, completed tasks, files processed |
Key Scripts
| Script | Purpose | Requires |
scripts/google_mail.py | Gmail read/draft/send (Google API) | GOOGLE_CLIENT_ID, GOOGLE_CLIENT_SECRET, GOOGLE_MAILBOX |
scripts/jobtread_pave.py | JobTread Pave API client | JOBTREAD_GRANT_KEY, JOBTREAD_ORG_ID |
scripts/drive_to_jobtread.py | Email→Drive→PM pipeline | Google Drive + JobTread credentials |
scripts/daily_briefing.py | Morning briefing generator | All of the above |
Adapting to Other PM Systems
If you use Procore, Buildertrend, CoConstruct, or another construction PM system instead of JobTread:
- Replace
scripts/jobtread_pave.pywith a client for your PM system's API - Update the query/mutation structure in
scripts/drive_to_jobtread.pyto use your PM system's file upload endpoint - Update
scripts/daily_briefing.pyto query your PM system's task/project schema - Update the environment variables to match your PM system's auth method
The pipeline pattern (Drive → download → upload → move → notify) stays the same.
Pitfalls
- PM system API keys can expire — JobTread grant keys expire after 3 months of inactivity. A daily briefing cron that queries the API keeps it active. If your PM system has similar expiration, keep it active with scheduled calls.
- Google OAuth tokens refresh automatically but can fail if the refresh token is revoked (password change, admin revocation). Monitor for auth errors and re-run the authentication flow when needed.
- Pave is NOT GraphQL — it's similar but different syntax. Inputs are passed after the
$symbol, and empty braces{}mean "return this field." Check the API explorer for your PM system before assuming GraphQL syntax works. - File routing is non-destructive — the pipeline moves Drive files to a "processed" folder rather than deleting them. This is intentional. If your pipeline deletes, add a confirmation step.
- Headless VPS auth — the one-time Google OAuth flow needs a browser. Use SSH port forwarding (
ssh -L 8080:localhost:8080) to complete the flow from your local machine, or run the auth on your local machine and copy the token file to the VPS. - Large file handling — construction documents (plans, permits) can be large. The Drive → PM upload uses a temp directory; ensure sufficient disk space. Clean up temp files after each run.
Verification Checklist
- Google Cloud Project created with Gmail, Drive, Calendar APIs enabled
- OAuth credentials created (Desktop app type)
- Environment variables set:
GOOGLE_CLIENT_ID,GOOGLE_CLIENT_SECRET,GOOGLE_MAILBOX - One-time OAuth authentication completed (token saved)
- Test:
google_mail.py read --count 5returns inbox - JobTread (or PM system) grant key obtained
- Environment variables set:
JOBTREAD_GRANT_KEY,JOBTREAD_ORG_ID - Test:
jobtread_pave.py projectsreturns project list - Drive folder IDs identified (raw + processed)
- Test:
drive_to_jobtread.py --dry-runshows pending files - Test:
drive_to_jobtread.pyprocesses one file end-to-end - Daily briefing cron configured and sending
- Drive poll cron configured and running
Related
- outlook-graph — Microsoft 365 email alternative (if the business uses Outlook instead of Google Workspace)
- crm-lite — lightweight CRM if the PM system doesn't handle lead tracking
- lead-to-payment — full sales pipeline orchestration
references/google-workspace-setup.md— step-by-step Google OAuth setupreferences/jobtread-api-setup.md— JobTread Pave API guide
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

