Client & Candidate Communication: Candidate Notes to Client-Facing Overview
Prompt: Candidate Notes → Client Facing Overview
Description: Reads every note on a person — across all their candidacies and their person record — and synthesizes them into a single, consistently formatted candidate overview written for a client audience. Applies the client-visibility filter so nothing internal-only leaks. The output is a paste-ready note (the connector can read notes but cannot write them back into Clockwork).
Frequency: Ad hoc (before sharing a candidate with a client, or building a slate)
Connectors: Clockwork Recruiting
Prompt: (copy/paste the prompt below into your chat in your LLM)
When I give you a person's name (and optionally the specific search this overview is for), build a client-facing Candidate Overview from their Clockwork notes. If the name is ambiguous, list matches and ask me to confirm.
Session setup: Load business-rules and domain knowledge if not already done this session, and cache statuses (list_reference_data(data_type=statuses)) and the firm-user map (for note attribution).
Step 1 — Gather every note. Pull the person record (get_person) and person notes (get_person_details(detail_type=notes, include=author)). Get all their candidacies via list_candidacies(person_id=...), then pull candidacy notes for each via get_candidacy_notes(include=author). Also pull positions, education, and compensation from get_person_details for factual grounding.
Step 2 — Apply the client-visibility filter (this is a client-facing document). Confirm once: "I'll include only notes and candidacies flagged visible to the client (stoplightStatus = green). Keep red and none excluded, or override for this run?" Then:
- Exclude any candidacy with stoplightStatus ≠ green.
- Exclude any individual note with stoplightStatus ≠ green, even on a green candidacy (filter notes independently).
- Exclude references/benchmarks (status.rank < 100).
- Check doNotContact — if true, flag it at the top and do not include outreach suggestions.
- If the source search is confidential (isConfidential = true), do not name the client company anywhere in the overview.
Step 3 — Synthesize, don't transcribe. Read across the filtered notes and resolve them into themes. Where notes conflict or a comp expectation changed over time, use the most recent and say so. Never invent detail that isn't supported by a note or a structured field; if a section has no shareable source material, write "Not yet captured" rather than guessing.
Step 4 — Output the overview using exactly this structure (this is the standard format — keep the section order and headers consistent every time so overviews are comparable across candidates):
[Candidate Name] [Current Title] [Current Company] [Location]
Summary — 2–3 sentence positioning statement: who they are and why they're a credible fit for this type of role.
Relevant Experience — 3–5 bullets of the most relevant accomplishments and scope, drawn from notes and work history.
Strengths — 3–4 bullets on what stands out, grounded in specific note observations.
Motivations & Fit — what's drawing them to a move / this opportunity, and any constraints (location, timing, role scope) that are shareable.
Compensation — current comp and expectations if captured in a green note or field; otherwise "Not yet captured."
Areas to Explore — 2–3 open questions or things still to validate (framed neutrally, client-appropriate — no internal-only concerns).
Recommended Next Step — one clear suggested action for the client.
Close with a one-line footer: how many notes were synthesized and how many were excluded by the visibility filter, so I know what's not shown. Output as clean Markdown I can copy into Clockwork or an email.
Adapt It: Swap in your firm's preferred section set or house style Produce an internal version (drop the green filter, add a candid "Recruiter Assessment / Risks" section) for team use Generate overviews for an entire slate in one pass so they read consistently side by side Tune length (a tight 150-word version for a slate cover page vs. a full one-pager) Feed an existing well-formatted overview as a model and have it match that structure instead of the standard one above