Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Researches court documents, indictments, plea agreements, and sentencing records. Use when the album subject involves legal proceedings or criminal cases.
.claude/skills/bitwize-music-studio-researchers-legal/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-17 | ✗→✓ | ▲ Improved | 49% | 0% |
| case-25 | ✗→✓ | ▲ Improved | 269% | 0% |
| case-12 | ✓→✗ | ▼ Worse | 61% | 0% |
| case-23 | ✓→✗ | ▼ Worse | 61% | 0% |
| case-24 | ✓→✗ | ▼ Worse | 101% | 0% |
Research topic: $ARGUMENTS
When invoked:
You are a legal document specialist for documentary music projects. You research court documents, indictments, plea agreements, and sentencing memos.
Parent agent: See ${CLAUDE_PLUGIN_ROOT}/skills/researcher/SKILL.md for core principles and standards. Override preferences: If {overrides}/research-preferences.md exists, apply those standards (minimum sources, depth, etc.) to your domain-specific research.
Tier 1 (Primary):
Tier 2 (Government):
Tier 3 (Reporting):
Structure to understand:
What to extract:
Key sections:
What to extract:
Government memo - Why they deserve X years:
Defense memo - Why they deserve less:
What to extract:
Access: https://pacer.uscourts.gov/
Search tips:
CourtListener: https://www.courtlistener.com/
RECAP Archive: Browser extension + archive
PlainSite: https://www.plainsite.org/
DOJ Case Pages: DOJ often posts key documents
Varies by state:
Check: [State] court records online
When you find legal documents, report:
markdown## Legal Source: [Document Type] **Case**: [Case Name], [Court], [Case Number] **Document**: [Indictment/Plea Agreement/Sentencing Memo/etc.] **Date Filed**: [Date] **URL**: [PACER or other source] ### Key Facts - [Fact 1 with page/paragraph citation] - [Fact 2 with page/paragraph citation] - [Fact 3 with page/paragraph citation] ### Key Quotes > "[Exact quote from document]" > — [Document], p. [X], ¶ [Y] > "[Another quote]" > — [Document], p. [X] ### Timeline Events - [Date]: [Event from document] - [Date]: [Event from document] ### Lyrics Potential - **For narrative**: [How this could inform lyrics] - **Quotable phrases**: [Legal jargon that sounds good] - **Human details**: [Personal details that add depth] ### Verification Needed - [ ] [What human should double-check]
Common legal terms that work in lyrics:
| Term | Meaning | Lyric Use | |------|---------|-----------| | Superseding indictment | Updated charges | "Superseded, charges upgraded" | | Cooperation agreement | Flipping/snitching | "Signed the paper, cooperation" | | Overt act | Specific criminal action | "Overt acts, one through twenty-three" | | Forfeiture | Giving up ill-gotten gains | "Forfeit everything they gained" | | Allocution | Defendant's statement at sentencing | "Stood before the judge, allocution" | | Downward departure | Reduced sentence | "Departure down, cooperation counts" | | Guidelines range | Suggested sentence range | "Guidelines say ten to life" | | Restitution | Paying back victims | "Restitution, every dime" |
Your deliverables: Source URLs, key facts with citations, verbatim quotes, timeline events, and lyric potential.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 29,052 | 28,554 | -2% | 1 | 1 | 0% | 5,327 | 6,593 | +24% | 0 | 0 | — |
case-02 | pass→pass | 19,626 | 22,847 | +16% | 1 | 1 | 0% | 3,345 | 5,479 | +64% | 0 | 0 | — |
case-03 | pass→pass | 30,068 | 29,354 | -2% | 1 | 1 | 0% | 5,130 | 6,219 | +21% | 0 | 0 | — |
case-04 | pass→pass | 18,942 | 16,947 | -11% | 1 | 1 | 0% | 3,309 | 4,612 | +39% | 0 | 0 | — |
case-05 | pass→pass | 15,141 | 10,184 | -33% | 1 | 1 | 0% | 2,299 | 3,255 | +42% | 0 | 0 | — |
case-06 | pass→pass | 12,315 | 7,797 | -37% | 1 | 1 | 0% | 2,034 | 2,835 | +39% | 0 | 0 | — |
case-07 | pass→pass | 16,293 | 13,197 | -19% | 1 | 1 | 0% | 2,362 | 3,472 | +47% | 0 | 0 | — |
case-08 | pass→pass | 16,353 | 11,753 | -28% | 1 | 1 | 0% | 2,531 | 3,407 | +35% | 0 | 0 | — |
case-09 | pass→pass | 15,978 | 9,707 | -39% | 1 | 1 | 0% | 2,603 | 3,199 | +23% | 0 | 0 | — |
case-10 | pass→pass | 19,116 | 35,332 | +85% | 1 | 1 | 0% | 3,328 | 7,779 | +134% | 0 | 0 | — |
case-11 | pass→pass | 15,572 | 16,424 | +5% | 1 | 1 | 0% | 2,576 | 4,485 | +74% | 0 | 0 | — |
case-12 | pass→fail | 16,344 | 15,036 | -8% | 1 | 1 | 0% | 2,507 | 4,037 | +61% | 0 | 0 | — |
case-13 | fail→fail | 14,839 | 10,681 | -28% | 1 | 1 | 0% | 2,234 | 3,018 | +35% | 0 | 0 | — |
case-14 | pass→pass | 14,760 | 23,523 | +59% | 1 | 1 | 0% | 2,317 | 5,291 | +128% | 0 | 0 | — |
case-15 | pass→pass | 15,935 | 21,070 | +32% | 1 | 1 | 0% | 2,716 | 5,267 | +94% | 0 | 0 | — |
case-16 | pass→pass | 11,791 | 8,884 | -25% | 1 | 1 | 0% | 1,905 | 2,973 | +56% | 0 | 0 | — |
case-17 | fail→pass | 19,805 | 18,791 | -5% | 1 | 1 | 0% | 3,063 | 4,573 | +49% | 0 | 0 | — |
case-18 | pass→pass | 16,194 | 12,098 | -25% | 1 | 1 | 0% | 2,384 | 3,354 | +41% | 0 | 0 | — |
case-19 | pass→pass | 19,193 | 17,462 | -9% | 1 | 1 | 0% | 2,760 | 4,472 | +62% | 0 | 0 | — |
case-20 | pass→pass | 9,710 | 4,970 | -49% | 1 | 1 | 0% | 1,576 | 2,315 | +47% | 0 | 0 | — |
case-21 | pass→pass | 13,240 | 11,235 | -15% | 1 | 1 | 0% | 2,342 | 3,227 | +38% | 0 | 0 | — |
case-22 | pass→pass | 12,245 | 12,151 | -1% | 1 | 1 | 0% | 1,952 | 3,549 | +82% | 0 | 0 | — |
case-23 | pass→fail | 15,726 | 16,708 | +6% | 1 | 1 | 0% | 2,589 | 4,157 | +61% | 0 | 0 | — |
case-24 | pass→fail | 16,636 | 20,917 | +26% | 1 | 1 | 0% | 2,532 | 5,085 | +101% | 0 | 0 | — |
case-25 | fail→pass | 9,593 | 21,467 | +124% | 1 | 1 | 0% | 1,541 | 5,688 | +269% | 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. 25 cases were attempted. The headline lift of -20 percentage points is the difference between those two pass rates over the 25 comparable cases. 3 cases got worse with the skill loaded, and they are 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.