ingest-profilelisted

# ingest-profile > **Runs under AGENTS.md.** These contracts bind without being restated here: Privacy Invariant (`current_base` never outbound), Honesty Firewall, Placeholder/Bracket Ban, Gate Write-back, Domain-Neutral Rule, Browser Automation Contract, Activity Pulse logging, Tracker verify+snapshot, and Sent-Clears-Draft. Inline reminders at point-of-use are intentional; standalone restatements point back to the relevant AGENTS.md section. > **Agent voice.** Read candidate modes through the shared DB-first accessor (`modes.agent_voice`, default `standard`) before producing any summary or explanation output. Apply the register semantics from AGENTS.md#mode-switches. This skill's interview is conversational by design, but **step confirmations, progress summaries, and section wrap-ups** must respect the register — e.g. `exec-summary` means a one-line "Got it — moving to targets" not a paragraph recap. Do not ask for a voice mode during initial setup; keep the stored value or use `standard`. > **Write mechanism depends on context.** This skill runs in two places, and only one has a shell: > - **One-shot CLI context** — an agent shell exists. Every `careerrat ...` command named throughout this file runs directly, as written. > - **Conversational chat context** (the onboarding screen, now the primary path) — tools are Read/Glob/Grep/Skill only. There is no shell, so every `careerrat ...` command in this file is unreachable. The only way to save anything is to emit a fenced c