| name | fclt-setup |
| description | Install, update, inspect, and initialize fclt from Codex. |
| tags | ["fclt","setup","codex","onboarding"] |
fclt-setup
When To Use
Use this skill when a user wants Codex to install, update, configure, inspect, or repair fclt.
Use it for:
- checking whether
fclt is installed and current
- initializing global
~/.ai or project <repo>/.ai
- installing or refreshing the built-in operating-model pack
- checking setup health with
doctor
- finding canonical, generated, runtime, and review paths
Workflow
- Bootstrap the complete loop with one idempotent command:
fclt setup
This initializes or safely updates global capability, initializes the current git repository
when present, prepares writeback/evolution review state, and installs the Codex plugin when Codex
is available.
- Inspect runtime selection and compatibility with
fclt_runtime action
status. Report the selected executable, version, source, protocol
compatibility, and fresh-session state.
If no compatible runtime is available, use the staged lifecycle:
check is read-only
stage requires an explicit version and approval, but does not activate it
apply requires approval plus the staged checksum precondition
rollback verifies and restores the retained prior runtime
Never curl-pipe code, use an unverified mutable URL, or replace an existing
global installation silently.
- Check current setup state and exact repair actions:
fclt --version
fclt paths --json
fclt doctor --json
Through MCP, call fclt_setup with an explicit global or
global_and_project scope. Project setup also requires the exact cwd.
Preview is the default; apply requires both dryRun: false and
approve: true.
- For advanced manual recovery, initialize global capability when missing:
fclt templates init operating-model --global
- If a repo needs local capability, initialize project AI:
fclt templates init project-ai
- Refresh pack defaults non-destructively:
fclt templates init operating-model --global --update --dry-run
fclt templates init operating-model --global --update
- Use
--force only when the user explicitly wants to replace local edits.
Rules
- Preserve existing
AGENTS.md, CLAUDE.md, and AGENTS.global.md guidance.
- First install should seed from existing agent guidance when available.
- Treat
doctor --json issues as setup facts, not user-facing blame.
- Treat Codex plugin registration as weaker evidence than fresh-session tool discovery.
- Treat external trackers as separate integrations. Core readiness depends only on configured local evidence coverage, not a vendor plugin or token.
- Prefer temp-root smoke tests for install/update behavior.
- Do not enable managed rendering unless the user wants fclt to write tool homes.
- Preview before mutation and state the exact global/project/plugin target.
- Do not report a staged runtime or installed plugin as active until the active
handshake and fresh-session discovery have been verified.
Output
- current installed version
- setup health
- paths that matter
- commands run
- what changed
- problem, evidence, reason, target, risk, and expected outcome
- verification performed and its actual result
- assumptions and fresh-session state
- exact undo or rollback path
- what still needs approval