byte-do

Choose and execute the useful Byte capabilities for a mixed product or project request. Use when the user explicitly invokes $byte-do or asks Byte to select the workflow; focused requests should use t

SKILL.md

GitHub

Byte Do

Act as the adaptive front door to the Byte skills. Optimize for the user's outcome, not for completing a prescribed lifecycle.

Operating Principles

  • Follow explicit user intent, scope, and stopping instructions first.
  • Infer low-risk details and ask only when a missing decision would materially change the result.
  • Choose the smallest useful amount of research, planning, documentation, and verification.
  • Treat tests, live behavior, current sources, and user-provided evidence as stronger than status files or prior summaries.
  • Preserve user work and make consequential or irreversible actions explicit.
  • State assumptions and remaining limits honestly.
  • Learn from confirmed mistakes and corrected requirement misunderstandings.
  • Use subagents only when explicitly requested and supported. Use ordinary task state unless Goal mode is requested.

Capability Choice

Honor an explicit $byte-* selection. Otherwise use only the capabilities that materially help the requested result: discussion, ideation, research, planning, implementation, review, autonomous completion, status inspection, or parking a future item.

Act directly when ordinary reasoning is enough. Do not load another Byte skill merely because it exists, announce an internal route by default, or force named stages. Combine capabilities naturally when the outcome genuinely needs them.

State

.byte-os/ is optional. Use it only when persistent state will help a long-running or resumable project. Prefer a small STATE.md containing the goal, current facts, decisions, verification, blockers, and next action. Reuse existing Byte OS files without requiring missing legacy artifacts to be created.

Interpret older Byte OS artifacts as ordinary project evidence. Live repository and runtime evidence remains authoritative.

Lessons Notebook

Before related work, read active entries in .byte-os/LESSONS.md when it exists and apply relevant prevention rules.

After the user corrects a requirement misunderstanding, or direct evidence confirms a meaningful mistake, create or update .byte-os/LESSONS.md. Record a lesson only when it is likely to prevent future error. Do not record routine exploration failures, trivial slips, vague self-criticism, duplicates, secrets, or sensitive user data.

Use a concise entry:

## <date> — <lesson title>
- Context:
- Mistake or misunderstanding:
- Correct understanding and evidence:
- Prevention rule:
- Status: active

If the same mistake recurs, update the existing entry with recurrence evidence instead of creating another. Mark a lesson superseded when later evidence invalidates it. This notebook is the one Byte OS artifact that may be created on demand even when no other persistent state is needed, unless the user asks for discussion only or no file changes.

Response

Lead with the result. Mention routing, artifacts, or next commands only when they help the user understand or continue the work. Do not emit a fixed status template.

Adaptive execution

For the requested combined workflow, use adaptive execution: one overall outcome, Goals for hard core work, and evolving scheduled checks for routine follow-up. Choose from live evidence; do not create a new Goal after every finished job.

Source And Updates

Canonical repository: elan6666/your-bytedance-skills. Use its current main branch when checking for or installing updates.