Deep interview
Clarify ambiguous tasks, inspect constraints, identify risks, and define acceptance criteria before code changes.
Documentation
Skills
Reusable Gajae Code workflows for interviewing, planning, durable goals, and goal-directed research.
Core workflow skills
The default product surface is four bundled workflow skills and four role agents — a method people can hold without overhead. Worker coordination is IRC, not a fifth skill. gjc team and gjc rlm are gone.
Ralplan
Turn a shaped task into an execution plan with critique points, validation commands, and success conditions.
Ultragoal
Drive implementation as a durable goal: act, check, revise, and report evidence instead of stopping at a patch.
Autoresearch
Run a goal-directed research mission that interleaves web research with the persistent Python REPL and ends on a structured verdict. Use it before an interview to map a landscape, or after a spec to gather evidence — never to implement.
Role agents
The matching four role agents stay read-only except executor:
Executor
Bounded implementation, fixes, and refactors.
Architect
Read-only architecture and code-review assessment.
Planner
Read-only sequencing and acceptance criteria.
Critic
Read-only plan critique and actionability review.
gjc-session helper
gjc-session provides scripts/gjc-session/create.sh for a human operator who needs a visible TUI with a named, worktree-backed session. It is not a fifth bundled workflow skill, not an automation interface, and does not grant automation authority.
For lifecycle control, task submission, status, and evidence boundaries, see Harness. Automated integrations use Coordinator MCP or the SDK; the visible gjc-session TUI remains human-operated.
Skill discipline
Skills are not magic prompts. They encode repeatable operating habits: context first, scoped authority, visible progress, and validation before completion.