Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Write internal company communications — 3P updates (Progress/Plans/Problems), company-wide newsletters, FAQ roundups, incident reports, leadership updates, status reports, project updates, and general internal comms. Use this skill any time the user asks to draft, edit, or format something meant for internal audiences. Trigger on keywords like "3P", "weekly update", "newsletter", "FAQ", "internal comms", "status report", "company update", "team update", "incident report", or any request to summa
.claude/skills/alirezarezvani-team-communications/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 50% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 209% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 43% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 75% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 79% | 0% |
> Originally contributed by maximcoding — enhanced and integrated by the claude-skills team.
Write polished internal communications by loading the right reference file, gathering context, and outputting in the company's exact format.
Identify the communication type from the user's request, then read the matching reference file before writing anything:
| Type | Trigger phrases | Reference file | |---|---|---| | 3P Update | "3P", "progress plans problems", "weekly team update", "what did we ship" | references/3p-updates.md | | Newsletter | "newsletter", "company update", "weekly/monthly roundup", "all-hands summary" | references/company-newsletter.md | | FAQ | "FAQ", "common questions", "what people are asking", "confusion around" | references/faq-answers.md | | General | anything internal that doesn't match above | references/general-comms.md |
If the type is ambiguous, ask one clarifying question — don't guess.
If the user hasn't connected Slack, Gmail, Drive, or Calendar, don't stall. Ask them to paste or describe what they want covered. You're formatting and sharpening — that's still valuable. Mention which tools would improve future drafts so they can connect them later.
| Anti-Pattern | Why It Fails | Better Approach | |---|---|---| | Writing updates without reading the reference template first | Output won't match company format — user has to reformat | Always load the matching reference file before drafting | | Inventing metrics or accomplishments | Internal comms must be factual — fabrication destroys trust | Only include data the user provided or MCP tools retrieved | | Using passive voice for accomplishments | "The feature was shipped" hides who did the work | "Team X shipped the feature" — active voice credits the team | | Writing walls of text for status updates | Leadership scans, doesn't read — key info gets buried | Lead with the headline, follow with 3-5 bullet points | | Sending without confirming audience | A team update reads differently from a company-wide newsletter | Always confirm: who will read this? |
| Skill | Relationship | |-------|-------------| | project-management/senior-pm | Broader PM scope — status reports feed into PM reporting | | project-management/meeting-analyzer | Meeting insights can feed into 3P updates and status reports | | project-management/confluence-expert | Publish comms as Confluence pages for permanent record | | marketing-skill/content-production | External comms — use for public-facing content, not internal |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-21 | pass→pass | 8,924 | 7,151 | -20% | 1 | 1 | 0% | 1,485 | 2,074 | +40% | 0 | 0 | — |
case-01 | fail→pass | 7,314 | 6,409 | -12% | 1 | 1 | 0% | 1,312 | 1,963 | +50% | 0 | 0 | — |
case-02 | fail→fail | 8,381 | 9,034 | +8% | 1 | 1 | 0% | 1,385 | 2,358 | +70% | 0 | 0 | — |
case-03 | fail→fail | 10,324 | 5,439 | -47% | 1 | 1 | 0% | 1,870 | 1,834 | -2% | 0 | 0 | — |
case-04 | pass→fail | 12,012 | 7,481 | -38% | 1 | 1 | 0% | 2,062 | 2,129 | +3% | 0 | 0 | — |
case-05 | pass→pass | 13,786 | 10,782 | -22% | 1 | 1 | 0% | 2,254 | 2,656 | +18% | 0 | 0 | — |
case-06 | pass→pass | 9,096 | 11,249 | +24% | 1 | 1 | 0% | 1,853 | 2,996 | +62% | 0 | 0 | — |
case-07 | fail→fail | 5,473 | 4,495 | -18% | 1 | 1 | 0% | 883 | 1,570 | +78% | 0 | 0 | — |
case-08 | fail→fail | 5,828 | 6,169 | +6% | 1 | 1 | 0% | 1,025 | 1,794 | +75% | 0 | 0 | — |
case-09 | fail→pass | 4,014 | 9,270 | +131% | 1 | 1 | 0% | 684 | 2,116 | +209% | 0 | 0 | — |
case-10 | pass→pass | 6,746 | 7,605 | +13% | 1 | 1 | 0% | 1,209 | 1,998 | +65% | 0 | 0 | — |
case-11 | fail→fail | 7,171 | 5,893 | -18% | 1 | 1 | 0% | 1,337 | 2,030 | +52% | 0 | 0 | — |
case-12 | fail→fail | 2,544 | 5,112 | +101% | 1 | 1 | 0% | 499 | 1,728 | +246% | 0 | 0 | — |
case-13 | fail→fail | 6,230 | 5,005 | -20% | 1 | 1 | 0% | 1,101 | 1,802 | +64% | 0 | 0 | — |
case-14 | fail→pass | 7,765 | 6,398 | -18% | 1 | 1 | 0% | 1,361 | 1,941 | +43% | 0 | 0 | — |
case-15 | fail→pass | 6,612 | 6,866 | +4% | 1 | 1 | 0% | 1,166 | 2,036 | +75% | 0 | 0 | — |
case-16 | fail→fail | 8,084 | 5,919 | -27% | 1 | 1 | 0% | 1,358 | 1,874 | +38% | 0 | 0 | — |
case-17 | fail→pass | 6,550 | 5,865 | -10% | 1 | 1 | 0% | 1,030 | 1,839 | +79% | 0 | 0 | — |
case-18 | pass→pass | 8,056 | 4,729 | -41% | 1 | 1 | 0% | 1,541 | 1,670 | +8% | 0 | 0 | — |
case-19 | pass→pass | 4,676 | 4,810 | +3% | 1 | 1 | 0% | 877 | 1,708 | +95% | 0 | 0 | — |
case-20 | pass→pass | 6,036 | 5,602 | -7% | 1 | 1 | 0% | 1,074 | 1,871 | +74% | 0 | 0 | — |
case-22 | fail→fail | 6,209 | 6,106 | -2% | 1 | 1 | 0% | 1,006 | 1,974 | +96% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +18 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.