# Claude Code Configuration

@../AGENTS.md

<!-- The import above (FLY-1268) is the single source of shared agent
     instructions — hard rules, workflow, operations, formatting. Claude Code
     inlines imported files at session start; the path is relative to this
     file, hence `../`. Do not restate AGENTS.md content here: every section
     below must be Claude Code-specific, and the drift-guard test
     (src/lib/agent-instructions.test.ts) fails on any heading this file
     shares with AGENTS.md. -->

## Slash Commands

<!-- Harness-specific: slash commands are a Claude Code feature (installed
     under .claude/commands/); other harnesses reach the same stage
     procedures by reading the skill files named in AGENTS.md. -->

Slash commands (`/capture`, `/refine`, `/activate`, `/implement`, `/review`,
`/validate`, `/close`, `/start-session`, `/wrap-session`, `/knowledge`,
`/generate-context`) load
the matching workflow procedure automatically. For workflow operations without
a slash command, prefer skill-led reasoning — read the relevant stage file
before acting. For general coding tasks, just write code.

## Hooks

<!-- Harness-specific: these are Claude Code hook lifecycle events, wired in
     .claude/settings.json and implemented in .claude/hooks/. Cursor has its
     own wiring in .cursor/hooks.json; other harnesses have none. -->

Hooks enforce the workflow automatically — formatting, transition validation,
session-file updates, and the Stop-gate lifecycle check. They fire on their
own; nothing to invoke. When a gate refuses something and you need the full
event table, read
`.claude/skills/flydocs-workflow/reference/claude-hooks.md`.