Normalize spend requests and prepare policy/evidence checks for human approval. The agent keeps the steps, checks and output consistent, while you keep every decision.
Picture tomorrow's handoff.
"Hello, review these new spend requests"
| Request | Amount | Policy check | Evidence | Approver | Status |
|---|---|---|---|---|---|
| Figma renewal (Design) | $540/yr | Under $1k software threshold - manager approval only | Invoice attached, matches last year's seat count | Priya (Eng manager) | Approved |
| New analytics tool | $4,800/yr | Over $1k - needs finance lead sign-off | Quote attached; no security review on file | Unassigned | Blocked - security review needed |
Sample data. Your agent fills the file from the sources you approve and flags anything it cannot verify.
Three files. One complete workflow.
skill.mdThe full job, evidence rules, output checks and stop points.
setup-prompt.txtCreates the tools, named Drive folder and sample-data test.
usage-prompt.txtThe short request you reuse whenever the job comes up.
Working in three steps
It checks access, creates the folder and builds the output file.
Use Google by default, or the Microsoft or Zoho alternative.
Unknowns and next decisions are shown instead of guessed.
Where the agent stops
- It does not invent missing facts or hide weak source evidence.
- It does not send, publish, spend, schedule or make a human decision without approval.
- It flags blocked work and tells you exactly what is needed next.
Put this job on a repeatable workflow.
Instant download · Keep the files forever · One-time $19
Add the next agent when this job is working
https://aiagentskills.biz/budgetvarianceagenthttps://aiagentskills.biz/cashflowforecastagent