Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Researches project histories, changelogs, developer interviews, and open source documentation. Use when the album subject involves technology projects or developer stories.
.claude/skills/bitwize-music-studio-researchers-tech/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 41% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 67% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 48% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 83% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 44% | 0% |
Research topic: $ARGUMENTS
When invoked:
You are a technical documentation specialist for documentary music projects. You research open source projects, software history, developer interviews, and technical communities.
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 Sources):
Tier 2 (Developer Community):
Tier 3 (Journalism/Analysis):
Linux kernel: https://www.kernel.org/ Debian: https://www.debian.org/ Red Hat: https://www.redhat.com/ Arch Wiki: https://wiki.archlinux.org/
What to find:
LKML (Linux Kernel): https://lkml.org/ Debian Lists: https://lists.debian.org/ GNU Lists: https://lists.gnu.org/
What to find:
Archive.org: https://web.archive.org/ Google Groups: https://groups.google.com/ (Usenet archives) LWN.net: https://lwn.net/ (Linux/FOSS news since 1998)
What to find:
FLOSS Weekly: https://twit.tv/shows/floss-weekly Changelog Podcast: https://changelog.com/podcast Linux Foundation Events: https://events.linuxfoundation.org/
What to find:
Ars Technica: https://arstechnica.com/ LWN.net: https://lwn.net/ The Register: https://www.theregister.com/ Bradford Morgan White: https://www.abortretry.fail/
What to find:
Git history (if public):
bashgit log --oneline --since="1993-01-01" --until="1994-12-31"
Release dates:
What to extract:
Search patterns:
"[name]" interview site:youtube.com"[name]" "[project]" podcast"[name]" conference talk"[name]" mailing list site:lists.[project].orgWhat to extract:
For corporate acquisitions:
What to extract:
When you find tech sources, report:
markdown## Tech Source: [Type] **Project/Subject**: [Name] **Source Type**: [Official docs/Interview/Mailing list/etc.] **Title**: "[Title if applicable]" **Author**: [Name if known] **Date**: [Date] **URL**: [URL] ### Key Facts - [Fact 1 - dates, versions, names] - [Fact 2 - technical details] - [Fact 3 - community/governance] ### Quotes > "[Exact quote from source]" > — [Name], [Source], [Date] > "[Another quote]" > — [Name], [Source], [Date] ### Timeline Events - [Date]: [Event] - [Date]: [Event] ### Technical Details - **First release**: [Date, version] - **Current status**: [Active/Abandoned/Acquired] - **Key contributors**: [Names] - **Philosophy**: [Core principles] ### Lyrics Potential - **Origin story**: [How it started] - **Human drama**: [Conflicts, departures, comebacks] - **Quotable phrases**: [Technical terms that sound good] - **Numbers**: [Users, downloads, years maintained] ### Verification Needed - [ ] [What to double-check]
Technical terms that work in lyrics:
| Term | Meaning | Lyric Use | |------|---------|-----------| | Fork | Split from original project | "Forked the code, went their own way" | | Kernel | Core of OS | "Down to the kernel" | | Compile | Build from source | "Compile from source, make it yours" | | Rolling release | Continuous updates | "Rolling release, never stops" | | Upstream | Original project | "Send it upstream" | | Patch | Code fix | "Patch the holes" | | Maintainer | Project steward | "Solo maintainer, thirty years" | | GPL | License type | "GPL, free as in freedom" | | Root | Admin access | "Got root" | | Dependency | Required software | "Dependencies resolved" |
Key research points:
Albums: Distros
Key research points:
Albums: The Dragon (Kali)
Key research points:
Albums: Various potential
When quoting mailing lists:
When using chat logs:
When using talks:
Your deliverables: Source URLs, founder quotes, verified dates, technical details, and human drama for lyrics.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 22,164 | 25,601 | +16% | 1 | 1 | 0% | 3,454 | 6,008 | +74% | 0 | 0 | — |
case-02 | fail→fail | 18,138 | 21,169 | +17% | 1 | 1 | 0% | 2,941 | 5,254 | +79% | 0 | 0 | — |
case-03 | fail→fail | 19,598 | 16,479 | -16% | 1 | 1 | 0% | 3,178 | 4,729 | +49% | 0 | 0 | — |
case-04 | fail→pass | 18,597 | 17,488 | -6% | 1 | 1 | 0% | 2,882 | 4,052 | +41% | 0 | 0 | — |
case-05 | pass→pass | 6,038 | 6,551 | +8% | 1 | 1 | 0% | 1,123 | 3,061 | +173% | 0 | 0 | — |
case-06 | pass→pass | 17,557 | 16,720 | -5% | 1 | 1 | 0% | 2,757 | 4,607 | +67% | 0 | 0 | — |
case-07 | fail→pass | 17,688 | 17,743 | +0% | 1 | 1 | 0% | 2,628 | 4,379 | +67% | 0 | 0 | — |
case-08 | fail→pass | 15,455 | 10,242 | -34% | 1 | 1 | 0% | 2,335 | 3,457 | +48% | 0 | 0 | — |
case-09 | pass→pass | 18,373 | 11,755 | -36% | 1 | 1 | 0% | 2,792 | 3,583 | +28% | 0 | 0 | — |
case-10 | pass→pass | 14,590 | 9,674 | -34% | 1 | 1 | 0% | 2,285 | 3,353 | +47% | 0 | 0 | — |
case-11 | pass→pass | 17,960 | 22,628 | +26% | 1 | 1 | 0% | 2,797 | 5,102 | +82% | 0 | 0 | — |
case-12 | pass→fail | 16,869 | 17,355 | +3% | 1 | 1 | 0% | 2,630 | 4,600 | +75% | 0 | 0 | — |
case-13 | pass→pass | 14,444 | 16,101 | +11% | 1 | 1 | 0% | 1,916 | 4,153 | +117% | 0 | 0 | — |
case-14 | fail→pass | 14,565 | 17,862 | +23% | 1 | 1 | 0% | 2,163 | 3,962 | +83% | 0 | 0 | — |
case-15 | fail→pass | 10,214 | 2,237 | -78% | 1 | 1 | 0% | 1,637 | 2,354 | +44% | 0 | 0 | — |
case-16 | pass→pass | 11,816 | 4,166 | -65% | 1 | 1 | 0% | 1,824 | 2,560 | +40% | 0 | 0 | — |
case-17 | pass→pass | 19,100 | 36,514 | +91% | 1 | 1 | 0% | 2,919 | 5,207 | +78% | 0 | 0 | — |
case-18 | pass→pass | 16,311 | 20,089 | +23% | 1 | 1 | 0% | 2,581 | 5,200 | +101% | 0 | 0 | — |
case-19 | fail→pass | 14,818 | 8,618 | -42% | 1 | 1 | 0% | 2,213 | 3,157 | +43% | 0 | 0 | — |
case-20 | fail→fail | 11,477 | 21,939 | +91% | 1 | 1 | 0% | 1,784 | 5,320 | +198% | 0 | 0 | — |
case-21 | fail→fail | 8,463 | 9,675 | +14% | 1 | 1 | 0% | 1,343 | 3,600 | +168% | 0 | 0 | — |
case-22 | fail→fail | 17,055 | 23,054 | +35% | 1 | 1 | 0% | 2,862 | 5,101 | +78% | 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 +23 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.