Output management
Save. Share. Verify. Reuse. The four verbs, every time.
Save the output where the team can find it next week: git, the team's KB (Notion / Confluence / SharePoint), or a prompts/ folder inside your Project. The chat scrollback is not durable storage.
Share with the person who can use it or review it, with the trail: the prompt, the source, and what you changed. A claim without a receipt is a guess with extra steps.
Verify before you count it as shipped — three checks in order: CI passes (before merge), stakeholder + legal + accessibility (before customer-facing), load tested + monitored + rollback plan named (before ship). Then reuse the pattern by extracting the prompt, building a template, and posting it to the team channel.
I'm a senior backend engineer on the Car-Mart API team (Python, 8 years, 2 repos in scope). I just generated a 400-word Claude summary of PR #1247's diff against our PR template. Walk me through Save, Share, Verify, Reuse for THIS specific output. For each verb: the concrete action, the tool (git, Slack, Notion, runbook repo, the Project's prompts/ folder, etc.), the success criterion I can check in under 30 seconds, and the one Car-Mart-shaped mistake to avoid.In your lane
Dev lane: Save the Claude summary as the PR description in PR #1247 (commit the file, link from the PR body). Share with the reviewer on Slack #api-refactor with the prompt + diff link. Verify Check 1 (CI passes, tests cover the change) and Check 2 (a peer reviewed, accessibility checked if UI). Reuse by saving the prompt to the api-refactor Project's prompts/ folder as 'pr-summary-template' and posting the link to #dev-tools. Concrete artifact: open Desktop → Project: api-refactor → new chat → paste PR #1247 diff → 'summarize per the PR template in CLAUDE.md, flag the missing Testing section' → save the prompt → run it on the next three PRs.
Data lane: Save the query to the analytics repo's queries/ folder with a README that names the data source and refresh cadence. Share the dashboard link in #data-team with the SQL + the stakeholder summary. Verify with the data steward (Check 2) and that dashboard freshness is monitored (Check 3). Reuse by saving the SQL-with-comments template to the q3-dashboard Project's prompts/ folder. Concrete artifact: open Desktop → Project: q3-dashboard → new chat → paste the SQL → 'explain this query, then list the 3 edge cases it does not cover, then format the result as a stakeholder summary' → save the prompt → run it on the next three queries.
Ops/Security lane: Save the runbook update to the ops repo's runbooks/ directory with the runbook link in the commit message. Share in #oncall with the on-call name and the escalation path. Verify with the on-call lead (Check 2) and the rollback plan is named in the runbook (Check 3). Reuse by saving the incident-triage prompt to the runbook-update Project's prompts/ folder. Concrete artifact: open Desktop → Project: runbook-update → new chat → drop in the new runbook → 'compare to the runbook template in CLAUDE.md, flag deltas in escalation path or on-call rotation' → save the prompt → run it on the next three runbook updates.
Pick one Claude output from today. Apply all four verbs in order. End state: the output is saved in its real home, shared with the trail, verified against the three checks, and the prompt is saved in the Project's prompts/ folder for next time.