Fde deploy solution
Privacy-safe forward-deployed engineering skills for Claude Code, Codex, and AGENTS.md-compatible agents
npx -y skills add inbusiness23/fde-skill --skill fde-deploy-solutionAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 16 days oldThe repository was created 16 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Test and introduce one approved workflow solution safely. Use after discovery and design to clean up prerequisites, test real historical cases, embed the result in existing tools, add logs and human approval, provide a manual fallback and off switch, and roll out to a small group before broader use.
SKILL.md
2.3 KB, 424 tokens by cl100k_base, as published. Nobody here has run it
Test and Introduce the Solution
Use plain language. Explain any necessary technical word immediately.
Start only when the better process, chosen fix, limits, owner, and cautious money case are approved.
Steps
- Fix the minimum important information, IDs, access, rules, and tool connections.
- List likely failures, who could be harmed, how to notice, when to stop, how staff continue manually, whether actions can be undone, and who responds.
- Gather old real cases with known answers, including normal, missing, confusing, old, conflicting, tricky, expensive, and must-go-to-a-person cases.
- Test each part, the whole process, retries, organization rules, and whether staff can use the answer.
- Put the answer in the CRM, store, job system, inbox, or approval screen employees already use. Keep that system as the official record.
- Build the smallest complete version: one start, minimum information, clear result, simple checks, human approval, work log, manual backup, and off switch.
- Test old cases, run silently beside staff, show staff recommendations, then try a small live group. Collect real-use results and run
fde-review-valuebefore any expansion. - Train staff, record approvals and corrections, and give the team the operating instructions and controls.
A person must approve important actions. Prices, payments, refunds, contracts, financing, schedules, private information, and customer promises require the full process and explicit live-use approval.
Read eval-embedding-rollout.md for required proof.
Return
Return cleanup completed, likely failures and controls, test results, where staff see the answer, current test group, monitoring and off switch, corrections, team readiness, and the next step. Return ready_for_value_review only after a controlled live test produces the required real-use results. Otherwise return blocked, followed by the exact missing test, control, approval, or evidence. Do not expand before fde-review-value returns expand.
What ships with it: 2 files
1.7 KB alongside SKILL.md
agents/
- openai.yaml218 B