Terminal
Terminal/Skill
Builds a recurring Clay workflow that flags when past buyers or customer contacts change jobs, with a follow and backfill play for each move.
By Clay
Works with
npx skills add clay-run/clay-skill-creator --skill track-champion-job-changesA champion who moves is the warmest pipeline you will ever get — someone who already bought or used your product, now sitting inside a new account with fresh budget authority and a honeymoon window in which tooling decisions are open. And every move is TWO plays, not one: the new company becomes a target (FOLLOW), and the vacated seat at your existing customer becomes a relationship risk (BACKFILL). Most teams run neither because nobody is watching. This skill builds the watcher: a scheduled Clay workflow that checks each champion's current employer, flags real moves, and produces a play-ready digest.
Clay workflows are an Alpha product — say so to the user before building, so they can calibrate expectations.
Nothing here ships with a value. Each one is the installer's, not the author's: ask for it, never substitute a plausible default, and where an answer does not exist say which step becomes unavailable rather than guessing. Where a default IS defensible it is named below, and using it means saying so in the output.
| Input | What the installer supplies | If it is missing |
|---|---|---|
| The champion list | where champions live today, and per row a name, a profile URL and the account they are known from | if there is no profile URL, an email works and gets resolved. A list that is not yet a re-checkable segment needs to become one, or nothing can run on a schedule |
| What counts as a champion | past buyer, power user, or deal contact on a closed-won | ask, and do not expand it silently — this defines the list |
| The ICP filter for the follow play | industry, size, region | ask. A mover whose new company sits outside ICP gets logged, not pursued |
| Cadence | how often the list is re-checked | weekly is defensible, monthly for smaller lists. State which |
| Destination | a table they read, a CSV, or a destination they own | the conversation is the fallback, and nothing is pushed anywhere they have not named |
If an answer sheet is present beside this skill, load it and ask only for what it does not cover. A partial sheet is normal; a value it is missing gets asked for on its own rather than restarting the interview. Say which values came from the sheet before using them — a sheet applied silently is a wrong field nobody catches. If there is no sheet, say nothing about sheets — the check is a file lookup, not a question, so run the interview as though the feature did not exist rather than reporting an absence. At delivery, offer to save the answers back (identifiers only — never a token or a password), private and never published — and phrase the offer so it explains itself: "want me to save your answers to a file, so the next person on your team doesn't have to answer these again?"
Run clay whoami; echo "exit_code=$?". If it fails, or the Clay workflow MCP tools
(read, edit_node, validate_workflow, execute_clay_action) are missing, run the Clay
plugin's setup skill (or follow
https://raw.githubusercontent.com/clay-run/agent-plugins/main/GETTING_STARTED.md), then
restart the agent if setup says to and re-run this skill. Confirm which workspace you are
signed into and tell the user before touching anything.
Present this plan, mapped to the user's inputs, and wait for approval before building
(follow the workflows skill's build protocol throughout):
TRIGGER: audience-scheduled over the "Champions" segment (weekly tick)
+ a manual trigger for testing
1. [tool] Resolve current employment. Resolution order matters:
(a) LinkedIn URL on file → person enrichment (cpj-enrich-person);
(b) no URL → people-index search by name + last-known company;
(c) reverse email→LinkedIn lookup LAST (weakest coverage; fails often).
→ current employer name, domain, title, current-role start date
2. [code] Extract + compare, deterministically — never an LLM. Pin the tool
node's whole $.result (deep paths fail the run when empty), find the
is_current experience entry, normalize both domains, and emit flat
string fields: verdict (current / moved / unverified), evidence with
dates, new-company domain/name/title/start. Use non-empty sentinels
("none") — a pinned value that resolves empty fails the run.
3. [conditional] Rules mode, routing on the verdict string (rules cannot compare two
dynamic fields — that is why the code node computes the verdict).
→ current: deterministic digest leaf (code node), end
→ unverified: "could not verify" digest leaf (code node), end
→ moved: continue. Genuinely ambiguous cases (rebrand/acquisition
suspicion) belong in the digest flagged for human review.
4a. FOLLOW branch (real move):
[tool] Enrich the NEW company (industry, headcount, region)
[conditional] ICP gate — outside ICP: log "moved, out of ICP", end
[agent] Compose the play: champion-arrival note anchored on the shared
history (which product, which account, when), plus 2-3 suggested
buying-committee titles to source at the new account
4b. BACKFILL branch (real move, runs in parallel):
[tool] Find people at the OLD account matching the vacated title/persona
[agent] Pick the most likely successor + note why; flag "seat vacated"
5. [leaf] Append one digest row per champion: verdict, evidence, plays
Build node-by-node with edit_node, confirm every action's real shape with
execute_clay_action before wiring it, run validate_workflow with prettier, and show the
user the graph. Where more than one Clay action can do a step (several person-enrichment or
people-finding functions usually exist), list the options by human-readable name with costs
and let the user choose.
Build gotchas verified against the live Alpha (2026-08):
def handler(context): returning a dict; read inputs with
context.get_input("name"). Top-level return is a syntax error.inputSchema shorthand ({"x": {"type":"string","sourceNodeId": "wfn_...","sourcePath":"$.field"}}). Pins that resolve to undefined OR empty string fail
the whole run — emit non-empty sentinels ("none") for possibly-empty fields.sourcePath: "$.result"
(the validator flags "$" with inputref_tool_node_path_missing_result_prefix); bind
it to a named input and dig in code. Deep paths from a CODE node resolve fine.pending forever and a downstream leaf runs on its empty output (silent blank results).
The verdict node merges resolve (prep→resolve→verdict, len 2) and the champion context
(prep→verdict, len 1) — that asymmetry hangs the run. FIX: put a passthrough "balancer"
code node on the short edge so both paths are equal length: prep→resolve→verdict AND
prep→balancer→verdict, balancer echoing champion_name + recorded_account_domain.
Verified: with the balancer, 3/3 ground-truth cases pass (current/moved/unverified),
1 credit/hit; without it, the verdict node never fires.{{vars}} on agent nodes fill reliably from flat string/object pins; a raw array
pin can leave the model claiming it got nothing. Flatten arrays in a code node first.clay workflows runs test — include at least one you
know has NOT moved (the no-change path must terminate cheaply) and, if possible, one
known mover.execute_clay_action / clay credits),
state the total, and get explicit approval.Per scheduled run, one digest with a row per champion:
champion · old account · verdict (current / moved / could-not-verify) · evidence · new company · new title · ICP fit · FOLLOW note draft · suggested committee titles · BACKFILL successor · seat-vacated flag
plus a summary line: champions checked, moves found, plays generated, rows needing human review.
Input: 120 champions in an Audience segment "Champions — Closed Won", weekly cadence, ICP = B2B software 100–5,000 employees, digest to a summary table. First test run of 5: 4 verified current, 1 real move — Jordan Lee, former Head of RevOps at longtime customer acme.example, now VP RevOps at northfield.example (2,300-person B2B software firm, in ICP). Digest row: verdict MOVED with LinkedIn evidence; FOLLOW note referencing the three years Jordan ran your product at Acme; suggested committee: CFO, Director of Sales Ops; BACKFILL: Priya Shah, current Director of RevOps at Acme, flagged as likely successor; seat-vacated alert on the Acme account. Weekly cost at 120 champions ≈ 150–350 credits; user approved before the first full run.