Task: Session Start Hook
Parent epic: #13 (Infrastructure)
Depends on: #21 (agent-brief.sh)
What to do
Create .claude/settings.json with a SessionStart hook that runs agent-brief.sh automatically when Claude Code starts in this repo. Every agent gets context before typing a single message.
{
"hooks": {
"SessionStart": [
{
"type": "command",
"command": "bash ./scripts/agent-brief.sh"
}
]
}
}
Why this matters
Without this, every agent session starts cold and must manually read CLAUDE.md + README + recent git log. With this, the brief runs automatically — zero token overhead for orientation, just immediate context.
Notes
- Only add the SessionStart hook — don't add other hooks that would interfere with agent work
- The hook output appears in the agent's context at session start
- If
agent-brief.sh doesn't exist yet, the hook should fail gracefully (not crash Claude Code)
- Keep
.claude/settings.json minimal — no settings that change agent behavior beyond the brief
Definition of done
Opening Claude Code in this repo shows the agent brief automatically in the first message context.
Task: Session Start Hook
Parent epic: #13 (Infrastructure)
Depends on: #21 (agent-brief.sh)
What to do
Create
.claude/settings.jsonwith a SessionStart hook that runsagent-brief.shautomatically when Claude Code starts in this repo. Every agent gets context before typing a single message.{ "hooks": { "SessionStart": [ { "type": "command", "command": "bash ./scripts/agent-brief.sh" } ] } }Why this matters
Without this, every agent session starts cold and must manually read CLAUDE.md + README + recent git log. With this, the brief runs automatically — zero token overhead for orientation, just immediate context.
Notes
agent-brief.shdoesn't exist yet, the hook should fail gracefully (not crash Claude Code).claude/settings.jsonminimal — no settings that change agent behavior beyond the briefDefinition of done
Opening Claude Code in this repo shows the agent brief automatically in the first message context.