K8s Exec

When to use which mode

Use psql when you need raw SQL — counting rows, joins, aggregations, schema inspection, CSV exports. It's faster and has output formatting options.

Use Django shell when you need the ORM, app models, Django utilities, or want to test Python logic against real data. Use it to verify code behavior, inspect model relationships, or run application-level checks.

Usage — psql mode

# Run a query
~/.claude/skills/k8s-exec/k8s_exec.sh -q "SELECT count(*) FROM some_table"

# Target a specific environment
~/.claude/skills/k8s-exec/k8s_exec.sh -n sandbox -q "SELECT count(*) FROM some_table"

# CSV output
~/.claude/skills/k8s-exec/k8s_exec.sh --csv -q "SELECT * FROM some_table LIMIT 10"

# Describe a table
~/.claude/skills/k8s-exec/k8s_exec.sh --describe some_table

# List databases
~/.claude/skills/k8s-exec/k8s_exec.sh --list-databases

# List pods
~/.claude/skills/k8s-exec/k8s_exec.sh --list-pods

# Interactive psql shell
~/.claude/skills/k8s-exec/k8s_exec.sh --shell

# SQL from file
~/.claude/skills/k8s-exec/k8s_exec.sh -f /path/to/query.sql

# Different database
~/.claude/skills/k8s-exec/k8s_exec.sh -d carta_web_audit -q "SELECT 1"

# Full options
~/.claude/skills/k8s-exec/k8s_exec.sh --help

Usage — Django shell mode

# Run Python code (ORM query)
~/.claude/skills/k8s-exec/k8s_exec.sh --python "from myapp.models import MyModel; print(MyModel.objects.count())"

# Multi-line Python (use quotes)
~/.claude/skills/k8s-exec/k8s_exec.sh --python "
from ce_migration.models import CEMigrationMetadata
qs = CEMigrationMetadata.objects.filter(status='completed')
for obj in qs[:10]:
    print(f'{obj.corporation_id}: {obj.carta_europe_id}')
"

# Python from file
~/.claude/skills/k8s-exec/k8s_exec.sh --python-file /path/to/script.py

# Python from stdin
echo "print('hello')" | ~/.claude/skills/k8s-exec/k8s_exec.sh --python-stdin

# Interactive Django shell
~/.claude/skills/k8s-exec/k8s_exec.sh --django-shell

# Target a specific environment
~/.claude/skills/k8s-exec/k8s_exec.sh -n sandbox --python "from django.conf import settings; print(settings.DATABASES.keys())"

Django shell safety

When using Django shell mode:

• Prefer read-only operations (queries, counts, prints) unless the user explicitly asks to modify data
• Avoid .save(), .delete(), or bulk updates unless explicitly requested
• Production is blocked by the script, but be cautious with preprod data too
• Test data modifications in sandbox first

Access errors on named environments

When targeting a well-known environment (sandbox, demo, preprod), kubectl commands may fail with permission/access/forbidden errors like:

• Error from server (Forbidden): pods is forbidden
• cannot list resource "pods" in API group
• Unauthorized

This means you don't have temporary access to the namespace. Automatically invoke the grant-k8s-access skill to request access, then retry the original command:

Use Skill tool: skill = "grant-k8s-access", args = "<ENVIRONMENT_NAME>"

Where <ENVIRONMENT_NAME> is the well-known environment name (e.g., sandbox). After the skill reports success, retry the original command.

How it works

The script auto-discovers the right pod:

• psql mode, dev namespaces: finds the postgres pod and runs psql directly (fast)
• psql mode, shared environments (no postgres pod): falls back to a carta-web app pod and uses manage.py dbshell
• Django shell mode: always uses a carta-web app pod (needs Django environment)