Panda Suite Docs
v1.9.2GitHub ↗
Panda Suite / Panda Lead

Panda Lead

Examples use standalone Codex names. For a Codex plugin, use $panda-suite:panda-lead; for a Claude Code plugin, use /panda-suite:panda-lead. See all invocation formats.

Panda Lead selects the smallest workflow that delivers your requested outcome. It coordinates; each specialist keeps its own method. Use it when a request spans several activities (discovery, planning, implementation, review, documentation) or when you are not sure which skill fits.

At a glance

Topic Summary
Use it for Compound requests, workflow choice, carrying context and authorization between activities
Not needed for A focused explanation, a tiny edit, or a task where you already know the specialist
Edits by default Only what the selected specialists are authorized to do
Main output Delivered artifacts plus an accurate status: implemented, verified, partially verified, blocked, or not started
Requires Nothing else; missing specialists are reported, never simulated

How Lead decides

  1. Reads the request. Lead identifies the outcome you want, the evidence available, your constraints, and the authorization you already gave. It inspects current artifacts before asking you for information. It matches the deliverable to your intent: an idea request gets exploration or a recommendation, not a finished artifact.
  2. Respects a named specialist. If you name a skill, Lead uses it. Otherwise it routes by the artifact you want, not by a keyword in your prompt.
  3. Loads only what it needs. Lead reads the selected skill through the host's discovery mechanism. Reading instructions is not a separate program running. If a skill is absent, Lead says so and explains what ordinary help it can still give.
  4. Orders dependent work. A feature build may need plan, then execute. An approved plan needs only execute. Review and documentation are added when you ask for them or the project's acceptance criteria require them. Lead carries a compact handoff between activities.
  5. Reports honestly. Lead compares the delivered work and actual checks with your requested outcome and names remaining decisions or risks.

Routing table

You want Lead selects
Compose named specialists under one shared outcome Craft
Explore an idea, assumptions, alternatives Brainstorm
Investigate a technical question or verify a claim Research
Learn an unfamiliar codebase Learn
Logos and visual identity Logo
Business rules, actors, acceptance Requirements
Tests: design, add, run, assess Test
What a proposed change could break Impact
Release readiness Release
Commits, conflicts, history Git
Create or evaluate Agent Skills Skill Lab
Initialize or prepare a project Setup
Ordered implementation tasks Plan
Implement authorized work Execute
Diagnose unexpected behavior Debugger
Performance and memory Perf
Dedicated security audit Security
Scanning, surface mapping, lab validation, bounty Red
Telemetry, incidents, fix verification, threat intelligence Blue
Defects in a change Review
Unnecessary complexity Simplify
Behavior-preserving restructuring Refactor
README, developer guide, runbook Write
Technical contract or design spec Spec
User journey instructions Manual

Buddhist Method is used only at a specific uncertainty, repeated failure, pressure point, or complexity decision, never as a stage between every activity.

Some overlaps are settled by rule: Requirements define what, Plan orders the work, Impact traces what may break. Test owns test effectiveness; Debugger owns causal investigation. Release preparation never implies deployment. Security handles dedicated audits while Review handles ordinary changed-code defects.

What to provide

  • The outcome you want and how you will judge it.
  • Relevant files, plans, URLs, or issue text.
  • Constraints such as compatibility promises or areas not to touch.
  • What may change: nothing, documents only, local code and tests, or more.

Examples

Full feature delivery from an approved plan

$panda-lead Implement the approved feature in docs/approval-plan.md, review the
change, add meaningful tests, and prepare release notes. Local code and docs may
change. Do not deploy, publish, push tags, or change production data.

Expect execute, review, test, and release-preparation evidence in that order, with no deployment. Planning is skipped because the plan already exists.

Read-only explanation

$panda-lead Explain how authentication works in this project.
Use the source as evidence. Do not change files.

A simple question can be answered directly. Lead may route to Learn for a guided walkthrough, but it will not edit anything.

From a vague idea to a plan, stopping before code

$panda-lead Customers keep asking for scheduled exports. Help me decide whether
we should build it, then turn the chosen direction into requirements and a plan.
Do not implement anything.

Expect brainstorm output with a recommendation, then requirements and a saved plan. Lead stops at the plan because implementation was not authorized.

Bug report to verified fix

$panda-lead Users report that saving a draft sometimes loses attachments.
Investigate and fix it, add a regression test, and write a short commit message.
Do not commit or push.

"Investigate and fix" authorizes the repair, so Debugger's fix proposal does not need a second approval. Expect a confirmed cause, the fix, an executed regression test, and a proposed commit message only.

Change assessment before a risky migration

$panda-lead We plan to split the users table into users and profiles. Tell me
what breaks, then plan the migration with release gates. Read-only except for the
saved plan.

Expect an Impact assessment followed by a Plan with release gates marked "defined; evidence pending".

Choosing a workflow only

$panda-lead I need to make our CSV import faster and also easier to maintain.
Which Panda skills should handle this and in what order? Do not start the work.

Expect a short routing recommendation (for example Perf for measured cost and Refactor or Simplify for structure) with reasons, and no work started.

Narrowing an earlier request

$panda-lead Change of plan: only review the current diff. Do not run the
release preparation we discussed earlier.

A new request can narrow earlier scope. Lead identifies the current outcome and carries forward only the evidence and authorization that still apply.

Authorization rules

  • Investigation, planning, review, and recommendations do not authorize source changes.
  • An explicit implementation request authorizes relevant reversible edits, and Lead carries that permission forward without asking again.
  • Destructive operations, publication, deployment, production changes, and external side effects need authorization that covers those actions.

Tips

  • Say what "done" means. Lead compares the result with it.
  • Name a specialist if you already know it; Lead will not override you.
  • You do not need Lead to use any other skill. Every specialist works alone.

Skill instructions

Read the full Panda Lead instructions in the repository.