Skip to content

fls-pilot logo

Safety & Limits

fls-pilot is rollback-first. Read-only checks can run freely, but persistent FL Studio project changes must be small, explicit, verified where FL Studio supports readback, recorded in the change log, and rollback-capable.

Normal Write Flow

For any workflow that might change the project, the assistant should:

  1. Scan first with read-only context.
  2. Explain the finding in normal user language.
  3. Propose one reversible next action with a risk level.
  4. Ask for explicit approval of that exact action.
  5. Apply at most one approved rollback unit.
  6. Read back the affected state where supported.
  7. Report before/after plus rollback or change_id, then stop.

Risk Levels

Level Meaning
read-only Reads project, server, or file context without changing FL Studio.
low One narrow rollback-backed write with clear target and restore data.
medium Wider rollback-backed write or state-sensitive target selection.
high Technically possible but too broad or weakly verifiable for default automation.
unsupported No safe FL Studio API path, no rollback/readback evidence, or a forbidden surface.

Hard Product Boundaries

fls-pilot does not expose tools for:

  • Loading or inserting plugins.
  • Rendering WAV/audio from FL Studio.
  • Project open, new, save-as, or broad UI automation.
  • Playlist clip placement, movement, splitting, or deletion.
  • Pattern or clip deletion.
  • Raw FL API escape hatches.
  • Full-FLP snapshot or full-project restore claims.

When FL Studio does not expose a safe API path, fls-pilot returns manual guidance or a read-only/probe-only result instead of pretending the assistant completed the action.

Knowledgebase-Backed Values

Plugin parameters, mixer ranges, dB/Hz mappings, REC event IDs, and indexing rules are not guessed. Runtime agents should use MCP resources and kb_* Knowledgebase tools for compact, verified context.