Skip to content

Developer Workflows

This page documents the practical workflows available through the agentic system — how to use them, what they do, and how to extend them.

Daily Briefing

Command: /gm or /briefing

The daily briefing gathers OSDU platform activity and combines it with your vault goals to produce a structured morning report.

What Happens

Step Action Source
1 Gather open MRs across monitored OSDU repos GitLab API via osdu-activity
2 Check pipeline status for recent failures GitLab API
3 Read current goals and key results Vault (01-goals/)
4 Read project monitoring config Vault (02-projects/)
5 Generate briefing note Wingman (briefing skill)
6 Write to vault brain skill → 00-inbox/YYYY-MM-DD.md
7 Print summary to terminal Wingman

Output

The briefing note includes:

  • Open MRs requiring attention (yours and others')
  • Pipeline failures or degraded services
  • Goal progress and upcoming deadlines
  • Suggested priorities for the day

Ship-It Workflow

Command: /send

Reviews, checks, commits, pushes, and creates a merge request in one flow.

What Happens

Phase Action Details
Preflight Check branch and working tree Ensure on a feature branch, no uncommitted conflicts
Review Lite code review Scan changes for obvious issues, security concerns
Quality Run quality checks terraform fmt -check, PowerShell syntax validation
Commit Generate commit message Via wt step commit (worktrunk) or aipr commit
Push Push to remote git push -u origin <branch>
MR Create merge request glab mr create --fill --target-branch dev

Prerequisites

  • On a feature branch (not dev or main)
  • Changes staged or unstaged (the workflow handles staging)
  • glab CLI authenticated to GitLab

Clone-and-Scan Workflow

Commands: /clone then /dep-scan

Clone OSDU repositories and analyze their dependencies.

Clone Phase

The /clone command sets up OSDU repos in .osdu/ using a bare clone + worktree layout:

.osdu/
├── os-partition/
│   ├── .git/              # Bare clone
│   └── master/            # Worktree (default branch)
├── os-entitlements/
│   ├── .git/
│   └── master/
└── ...

Available categories:

Category Repos
libraries os-core-common, os-core-lib-azure, os-oqm
core os-partition, os-entitlements, os-legal, os-storage, os-search, os-indexer
reference os-wellbore, os-well-delivery, os-seismic
ingestion os-ingestion-workflow, os-csv-parser, os-manifest-ingestion
domain os-reservoir, os-well-planning, os-dataset

Scan Phase

The /dep-scan command runs a comprehensive dependency analysis:

  1. Discover — Find all dependency manifests (pom.xml, package.json, requirements.txt)
  2. Fetch — Run Trivy vulnerability scan, parse dependency trees
  3. Assess — Score each dependency by risk (version age, CVEs, category)
  4. Distill — Generate summary with recommendations
  5. Write — Store report in vault (04-reports/)

Debug-and-Verify Workflow

Commands: /debug then /verify

Systematic root cause analysis followed by evidence-based completion verification.

Debug Phase

The /debug command triggers the iac-debug skill's four-phase investigation:

Phase Action
Gather evidence Read logs, events, pod status, recent deployments
Check recent changes git log, terraform state, helm history
Form hypothesis Correlate evidence with known failure patterns
Implement fix Apply the fix, test, document

Verify Phase

The /verify command triggers the iac-verify skill, which refuses to mark work complete without running verification commands:

Check Command
Cluster health kubectl get nodes, kubectl get pods -A
AKS safeguards kubectl get constrainttemplates
Elasticsearch curl elasticsearch-es-http:9200/_cluster/health
PostgreSQL kubectl cnpg status postgresql
cert-manager kubectl get certificates -A
Istio kubectl get peerauthentication -A

Workspace Layout

The project uses git worktrees managed by worktrunk (wt) for parallel development.

Branching Model

main ← dev ← feature/*
  • main — clean deployable surface (read-only reference)
  • dev — integration branch (daily work)
  • feature/* — individual changes, branched from dev

Worktree Structure

cimpl-azure-provisioning/          # Bare repo root
├── main/                          # main branch worktree
├── dev/                           # dev branch worktree (primary)
├── feature/docs-restructure/      # Feature branch worktree
└── .osdu/                         # OSDU service repos (bare + worktree)

Worktrunk Configuration

Defined in .config/wt.toml:

default-branch = "dev"              # Feature branches start from dev
post-create = "wt step copy-ignored --from dev"  # Copy .env files to new worktrees

[ci]
platform = "gitlab"

Common Commands

# Create a feature branch
wt switch --create feature/<name> --base dev

# See all changes since branching
wt step diff -- --stat

# Commit with generated message
wt step commit

# Push and create MR
git push -u origin $(git branch --show-current)
glab mr create --fill --target-branch dev

Extending the System

Adding a New Skill

  1. Create SKILL.md in the appropriate location:

    • Shared: .github/skills/<name>/SKILL.md
    • Wingman: .ai-wingman/skills/<name>/SKILL.md
    • OSDU: .ai-osdu/skills/<name>/SKILL.md
  2. Define the skill with:

    # Skill Name
    
    ## Triggers
    - keyword1, keyword2, phrase
    
    ## Protocol
    1. Step one
    2. Step two
    
    ## Constraints
    - Rule one
    - Rule two
    

  3. Add routing entry in .github/routing.md

  4. Add trigger mapping in .claude/CLAUDE.md skill loading table

Adding a New Slash Command

Create a file in .claude/commands/<name>.md with the command protocol. The filename becomes the command (e.g., deploy.md/deploy).

Adding a New Prompt

Create a file in .github/prompts/<name>.prompt.md. Prompts are longer multi-step workflows that can be invoked by commands or agents.

Adding a Vault Template

Create a template in .ai-brain/templates/<name>.md with YAML frontmatter and Obsidian-compatible markdown. Templates are used by the brain skill when creating new notes.


See also

  • Agents — agent profiles and delegation model
  • Skills & Commands — the skills and commands these workflows use
  • Knowledge Vault — the persistent memory layer
  • Design — the infrastructure and platform these workflows operate on