Blog
New perspectives, new possibilities.
87 results
Honest Reports From Multi-Agent Claude Code Reviews
One risk unverified, one test blocked, and the main agent says "all passed." Split results into confirmed, to verify, and not run, and mark missing state unknown.
2026-11-13 ·DeepDojo.AI
Grow Your Team's Claude Code Setup From One Command
Ship the full setup on day one and nothing runs; write one toolchain into commands, hooks, and CLAUDE.md and they clash. Start small, keep one source per job.
2026-11-13 ·DeepDojo.AI
Claude Code Plugin Upgrades You Can Roll Back
A new plugin version migrated its cache, and rolling back still crashed. Track secrets to every use, keep runtime data persistent, and back up before upgrading.
2026-11-13 ·DeepDojo.AI
When to Package Your Claude Code Setup as a Plugin
Copy configs by hand and every teammate drifts; package too early and each fix needs a team-wide update. Let it mature in the project, then package one scenario.
2026-11-13 ·DeepDojo.AI
Running Claude Code Subagents in Parallel Safely
Two subagents had separate worktrees but shared one database and overwrote each other's test data. Separate context, files, and resources before running in parallel.
2026-11-13 ·DeepDojo.AI
How to Brief a Claude Code Subagent
"Look at auth" comes back as "review complete," which the main agent can't check. Brief subagents on job, scope, tools, and output; keep facts apart from guesses.
2026-11-13 ·DeepDojo.AI
Stop Claude Code From Writing to Your Database
"Check the failed orders" became an UPDATE, and a read-write account let it through. Limit MCP with read-only accounts, aggregate at the source, vet outside text.
2026-11-13 ·DeepDojo.AI
When Claude Code Actually Needs an MCP Server
Remote PR comments and live database state aren't in local files, and careless MCP config leaks tokens. Decide with four questions, then pick transport and scope.
2026-11-13 ·DeepDojo.AI
Why Didn't My Claude Code Hook Block That?
A hook script missing jq exits with 127, and the risky call goes ahead. Separate refusals from script errors, rely on real boundaries, and test hooks in two steps.
2026-11-13 ·DeepDojo.AI
Which Event Should a Claude Code Hook Use?
A blocking script on PostToolUse fires after the files are gone. Choose the Claude Code hook event by what the check must change, then narrow it with matcher and if.
2026-11-13 ·DeepDojo.AI