Profiles and output styles

Setting	Controls
Profile	Visible tools, agent instructions and permission rules
Output style	Reply length, tone and explanation

The two settings can be combined independently.

Profiles

Built-in profiles

Profile	Available work
default	General agent execution
auto	Preapproved routine edits and shell commands; hard denials still apply
plan	Research and writes inside .gdc/plans/
review	Investigation and review notes; no shell execution or delegation
test	Read files and execute tests; no edits or delegation
migrate	Editing and shell execution; delegation denied
planner	Plan swarm goals; available to root sessions and workers
swarm-worker	Execute swarm assignments; available to root sessions and workers
coordinator	Delegate, inspect and combine worker results; no direct edits or shell execution

Subagent-only profiles include explore and verification. The hidden meta profile is reserved for internal operations.

Selecting a profile

/profile review

Shift+Tab cycles profiles in the terminal. For headless execution:

gdc print --profile plan "plan the auth refactor"

Your own profiles

Place Markdown profiles under ~/.gdc/agents/<name>.md or <project>/.gdc/agents/<name>.md. Configure extra directories in [agents]. Custom profiles can receive permission overrides. Built-in profiles ignore those overrides.

Output styles

Style	Reply format
concise	Short answers
verbose	Detailed reasoning and trade-offs
explanatory	Definitions and explanations of decisions

Selecting a style

gdc print --profile plan --output-style verbose "plan the migration"

[output_styles]
default = "concise"

Custom styles use Markdown files with name and description frontmatter under ~/.gdc/output_styles/ or project .gdc/output_styles/. The body supplies the style instructions.