agent-workflow
Six agents, two branches
The reasoning behind the devcontainer firewall, MCP wiring, CLI lifecycle, SSH signing, and build ordering lived only in scattered commit messages. A wayfinder map turned it into five permanent ADRs, produced by six agents working in true parallel.
The links belonged outside the prose
Two quick publication handoffs turn author feedback into a stricter rule for what belongs in a development post.
Collapsing the three-file translation triplet into one module
Maintaining UI translations across three separate files created silent sync friction. Moving to a single per-key module enforced compile-time translation parity.
The icon nobody recognized
Player feedback from an iOS user led to updating the backspace key icon and moving it to the top row.
Seven phases under one umbrella
A complete visual redesign across 18 catalogued surfaces, guided by a token migration plan, oklch contrast math, and a seven-phase base branch umbrella.
Done is not committed
Context7 and browser automation reach all three developer CLIs, after live verification exposes two integrations lost with an ephemeral worktree.
One skill, three agents
Moving the repository-owned skills to one provider-neutral home, restoring the Codex workspace sandbox, and making portability a tested claim.
The allowlist was too generous
Two routine blog handoffs expose a less routine problem: broad agent permissions were approving destructive commands without review.
More than one agent in the workshop
Codex and Antigravity join Claude Code inside the devcontainer, with pinned versions, shared skills, and checks that the safety boundaries are real.
Tuning the agent workbench: from game log polish to MCP servers
A solo-plus-agents project spends real time on its own tooling. Three days of making the workshop pleasant.