livepilot-sample-engine — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited livepilot-sample-engine (Agent Skill) and scored it 96/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 1 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 1 flagged
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
The Sample Engine is LivePilot's intelligence layer for sample discovery, analysis, critique, and creative manipulation. It turns raw samples into musical elements through 29 techniques drawn from Dilla, Burial, Amon Tobin, Four Tet, and DJ Premier.
Every sample workflow is guided by one of two philosophies:
seamlessly. The sample should sound like it was always part of the track.
The original creator shouldn't recognize their sample.
Stuck and need surprise? Alchemist. The critics and intent determine which.
| Tool | Purpose |
|---|---|
analyze_sample | Build SampleProfile — material type, key, BPM, recommendations |
evaluate_sample_fit | 6-critic battery — key, tempo, frequency, role, vibe, intent fit |
search_samples | Search Splice library, Ableton browser, and local filesystem |
suggest_sample_technique | Recommend techniques from the 29-recipe library |
plan_sample_workflow | End-to-end plan: analyze + critique + technique + compiled steps |
get_sample_opportunities | Analyze song for where samples could improve it |
User asks to do something specific with a sample:
"Chop this vocal into a rhythm"
-> analyze_sample -> evaluate_sample_fit(intent="rhythm")
-> suggest_sample_technique -> executeUser wants to find and use a sample:
"Find me a dark vocal for this track"
-> search_samples(query="dark vocal") -> present candidates
-> user picks -> analyze -> critique -> plan -> executeWhen stuck, Wonder Mode can suggest sample-based variants:
file_path from prior search_samples resultssearch_samples — the agent must chain the callsanalyze_sample tells you whatthe material is before you commit to a technique
evaluate_sample_fit catcheskey clashes, tempo mismatches, and frequency masking before they happen
even for the same sample
a proven recipe exists in the library
| Type | Detection | Best Simpler Mode | Best Warp Mode |
|---|---|---|---|
| vocal | "vocal", "vox", "voice" in name | Slice (Region) | Complex Pro |
| drum_loop | "drum", "break", "beat" in name | Slice (Transient) | Beats |
| instrument_loop | "guitar", "piano", "synth" | Slice (Beat) | Complex Pro |
| one_shot | "kick", "snare", "clap", short | Classic | Complex |
| texture | "ambient", "pad", "drone" | Classic | Texture |
| foley | "foley", "field", "recording" | Classic | Texture |
| fx | "fx", "riser", "sweep" | Classic | Complex |
| full_mix | full mix, long duration | Slice (Beat) | Complex Pro |
Each scores 0.0-1.0 on one dimension of fitness:
Simpler is the default sample container, but the installed packs ship instruments that change the physics of what playback means. When the user reaches for "more texture" / "less recognizable" / "evolving" / "grain" / "drone bed", consider these before defaulting to Simpler + warp:
These are awareness, not prescription — pick the one that fits the moment. Don't lock in a chain.
references/sample-techniques.md — Full 29-technique catalogreferences/sample-critics.md — Critic scoring detailsreferences/sample-philosophy.md — Surgeon vs Alchemist guidedocs/research/2026-04-22-ableton-packs-deep-analysis.md — Per-pack capabilities (44 installed packs scored against minimal/dub/cinematic aesthetic)~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.