Getting Started: How to Use This Prompt Playbook

How To Use This Prompt Playbook

Treat each prompt as a starting point. The “Adapt It” note for each prompt lists additional ways or suggestions to further refine and retarget the prompt. The more specific you are about project names, time windows, and tone, the better the output.

Start every session within your LLM by loading the knowledge base. The first Clockwork tool call should be:

clockwork_get_knowledge(topic='business-rules') then clockwork_get_knowledge(topic='domain')

Prompts with a "Session setup" block include this as step 0; for others it happens automatically via the tool's built-in reminder. Load once per session — the content is stable.

Establish your identity once. For "my projects" / "my candidates" prompts, resolve who you are at session start via clockwork_get_current_user (or set CLOCKWORK_CURRENT_USER_ID in your system prompt). Then list_projects (user_id=<me>) and get_user_projects return your book in a single call.

For recurring prompts (Daily, Weekly, Monthly), consider scheduling them or creating a skill in your LLM so the output is waiting in your inbox or Slack each morning.

Review the first few runs of any new prompt closely and tune the output format to your taste — small wording changes compound over time, and you should always check AI’s work.

On client-facing outputs: any prompt that generates client content will confirm the stoplightStatus visibility filter before running. Answer once per session — your answer carries through that run and is re-confirmed next time.

On project display: all prompts use the <Company> \ <Project> format (e.g., "Mission \ VP of Finance"), matching the Clockwork UI. Confidential searches display as [Confidential] \ <Role Title> in internal contexts.

On response format: Clockwork tools support response_format='json' for machine-readable output. For prompts that build lookup maps, compute aggregates, or feed multi-step calculations, request json on those calls. For final outputs you're reading directly, the default markdown is fine.

Mix and match. Many prompts chain naturally — a conflict check before an outreach sequence, a market map feeding a kickoff brief, or a completeness audit before a client update. The tools are building blocks; combine them to fit your workflow.

Use and load branded templates for reports. When running prompts to deliver client-facing deliverables, upload your firm’s template for the specific report when prompting.

Did this answer your question? Thanks for the feedback There was a problem submitting your feedback. Please try again later.