Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Summarize and analyze earnings calls (法說會) including financial highlights, management commentary, guidance, and analyst Q&A key takeaways. Use this skill when the user needs to digest an earnings call transcript, extract key financial data and forward guidance, compare actual results vs consensus estimates, or prepare for an investor meeting — even if they say 'summarize this earnings call', 'what did management say about next quarter', 'did they beat estimates', or 'key takeaways from the 法說會'.
.claude/skills/asgard-ai-platform-fin-earnings-summary/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 26% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 52% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 22% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-01 | ✓→✓ | = Same ✓ | -33% | 0% |
IRON LAW: Separate Facts from Spin
Management presentations emphasize positives and downplay negatives.
Your job is to extract: (1) hard numbers, (2) forward guidance with
specificity level, and (3) what management DIDN'T say (notable omissions).
"Strong momentum" without numbers is spin.
"Revenue grew 15% to NT$2.3B" is a fact.| Section | Duration | What to Extract | |---------|----------|----------------| | Prepared Remarks | 15-20 min | Revenue, EPS, margins, key metrics — actual vs guidance vs consensus | | Management Commentary | 10-15 min | Forward guidance, strategic initiatives, market outlook | | Q&A | 20-30 min | What analysts are worried about (revealed by their questions), management's candor level |
Financial Highlights (vs Consensus) | Metric | Actual | Consensus | Beat/Miss | YoY Change | |--------|--------|----------|-----------|-----------| | Revenue | ${X} | ${X} | Beat/Miss by {%} | {%} | | EPS | ${X} | ${X} | Beat/Miss by {%} | {%} | | Gross Margin | {%} | {%} | ±{pp} | ±{pp} | | Operating Margin | {%} | {%} | ±{pp} | ±{pp} |
Forward Guidance | Metric | Guidance | Prior Guidance | Change | Specificity | |--------|---------|---------------|--------|------------| | Next Q Revenue | ${X}-${X} | ${X}-${X} | Raised/Maintained/Lowered | Specific/Vague | | Full Year EPS | ${X}-${X} | ... | ... | ... |
Qualitative Signals
Phase 1: Extract Numbers
Phase 2: Evaluate Guidance
Phase 3: Read the Q&A
Phase 4: Compare to Market Reaction
markdown# Earnings Summary: {Company} — {Quarter} ## Headline {One sentence: beat/miss + most important takeaway} ## Financial Highlights | Metric | Actual | Consensus | vs Consensus | YoY | |--------|--------|----------|-------------|-----| | Revenue | ... | ... | ... | ... | | EPS | ... | ... | ... | ... | ## Forward Guidance | Metric | New Guidance | Prior | Change | |--------|------------|-------|--------| | ... | ... | ... | Raised/Maintained/Lowered | ## Management Commentary (Key Points) 1. {key point with quote} 2. {key point} ## Q&A Highlights - **Analyst concern**: {topic} → **Management response**: {summary} - **Notable omission**: {what wasn't addressed} ## Market Reaction - Post-earnings move: {+/-X%} - Interpretation: {why the market reacted this way} ## Investment Implication {What this means for the investment thesis — bullish/neutral/bearish signal}
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-03 | fail→pass | 23,835 | 18,512 | -22% | 1 | 1 | 0% | 3,541 | 4,473 | +26% | 0 | 0 | — |
case-01 | pass→pass | 31,898 | 17,706 | -44% | 1 | 1 | 0% | 6,449 | 4,291 | -33% | 0 | 0 | — |
case-02 | fail→pass | 25,125 | 29,281 | +17% | 1 | 1 | 0% | 3,576 | 5,433 | +52% | 0 | 0 | — |
case-04 | pass→pass | 18,205 | 12,758 | -30% | 1 | 1 | 0% | 2,327 | 3,079 | +32% | 0 | 0 | — |
case-05 | pass→pass | 14,381 | 12,738 | -11% | 1 | 1 | 0% | 2,083 | 3,063 | +47% | 0 | 0 | — |
case-06 | pass→pass | 11,139 | 6,759 | -39% | 1 | 1 | 0% | 1,621 | 2,145 | +32% | 0 | 0 | — |
case-07 | fail→pass | 14,322 | 8,067 | -44% | 1 | 1 | 0% | 1,696 | 2,065 | +22% | 0 | 0 | — |
case-08 | pass→pass | 20,922 | 25,398 | +21% | 1 | 1 | 0% | 2,654 | 4,287 | +62% | 0 | 0 | — |
case-17 | pass→pass | 18,862 | 18,337 | -3% | 1 | 1 | 0% | 2,530 | 3,344 | +32% | 0 | 0 | — |
case-09 | pass→pass | 17,953 | 16,063 | -11% | 1 | 1 | 0% | 2,350 | 3,447 | +47% | 0 | 0 | — |
case-10 | pass→pass | 14,792 | 12,438 | -16% | 1 | 1 | 0% | 1,797 | 2,794 | +55% | 0 | 0 | — |
case-11 | pass→pass | 16,181 | 14,708 | -9% | 1 | 1 | 0% | 2,025 | 2,762 | +36% | 0 | 0 | — |
case-12 | pass→pass | 11,662 | 10,664 | -9% | 1 | 1 | 0% | 1,591 | 2,518 | +58% | 0 | 0 | — |
case-22 | pass→pass | 35,904 | 28,876 | -20% | 1 | 1 | 0% | 8,268 | 7,466 | -10% | 0 | 0 | — |
case-13 | pass→pass | 17,927 | 10,234 | -43% | 1 | 1 | 0% | 2,570 | 2,662 | +4% | 0 | 0 | — |
case-14 | pass→pass | 11,067 | 10,037 | -9% | 1 | 1 | 0% | 1,720 | 2,630 | +53% | 0 | 0 | — |
case-15 | pass→pass | 17,324 | 11,975 | -31% | 1 | 1 | 0% | 2,330 | 3,083 | +32% | 0 | 0 | — |
case-16 | pass→pass | 11,433 | 5,045 | -56% | 1 | 1 | 0% | 2,035 | 2,066 | +2% | 0 | 0 | — |
case-18 | fail→pass | 14,143 | 12,544 | -11% | 1 | 1 | 0% | 2,109 | 2,938 | +39% | 0 | 0 | — |
case-19 | pass→pass | 15,829 | 8,835 | -44% | 1 | 1 | 0% | 2,004 | 2,416 | +21% | 0 | 0 | — |
case-20 | pass→pass | 14,061 | 9,395 | -33% | 1 | 1 | 0% | 1,900 | 2,214 | +17% | 0 | 0 | — |
case-21 | pass→pass | 9,885 | 5,944 | -40% | 1 | 1 | 0% | 1,851 | 2,191 | +18% | 0 | 0 | — |
case-23 | pass→pass | 25,233 | 16,466 | -35% | 1 | 1 | 0% | 3,589 | 3,984 | +11% | 0 | 0 | — |
case-24 | pass→pass | 40,507 | 18,005 | -56% | 1 | 1 | 0% | 3,766 | 4,842 | +29% | 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. 24 cases were attempted. The headline lift of +17 percentage points is the difference between those two pass rates over the 24 comparable cases.
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.