coreyhaines31/makerskills / skills/castmate
castmate
FlaggedWhen you want to pull in the right AI model for the right job: hand a task to another agent CLI (Codex, Claude, Gemini, Grok, Amp, OpenCode, Cursor, Kiro, Antigravity), or send one prompt to several at once for independent opinions, then synthesize. Delegate mode gives one castmate the job, optionally with write access ('have codex write the migration'). Fan-out mode runs read-only reviews in parallel. Rounds mode lets castmates build on and challenge each other's findings. Calls the official CLIs you already have installed and logged in (Gemini CLI needs an API key on personal accounts); no MCP. Triggers on '/castmate,' 'ask codex,' 'have gemini do this,' 'get a second opinion,' 'what would other models say,' 'fan this out,' 'cross-check this with another model,' 'delegate this to codex.' Advisor personas for founder questions go to maker-council. Personal models, casting rules, and run outputs live in ~/.config/makerskills/castmate/.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 12.7 KB |
| references/amp-readonly-settings.json | config | 229 B |
| references/casting.md | markdown | 2.6 KB |
| references/cli-setup.md | markdown | 11.3 KB |
| references/gemini-readonly-policy.toml | config | 857 B |
| references/presets/README.md | markdown | 757 B |
| references/presets/bughunt.md | markdown | 1.6 KB |
| references/presets/contracts.md | markdown | 1.1 KB |
| references/presets/hotspots.md | markdown | 1.7 KB |
| references/presets/invariants.md | markdown | 1.5 KB |
| references/presets/regression.md | markdown | 1.1 KB |
| references/presets/security.md | markdown | 1.5 KB |
| references/prompt-template.md | markdown | 3.1 KB |
| references/tools.json | config | 9.1 KB |
| scripts/castmate.pyexecutable | script | 21.1 KB |
| scripts/test_castmate.py | script | 11.7 KB |
references/prompt-template.md
View on GitHub# Prompt templates
Castmates run in the same working directory as the host and can read files themselves. Point them to files by path instead of pasting the contents; inline only a short snippet that frames the question, like an error message or a function signature.
**Inline diffs; don't ask castmates to run git.** Claude castmates have no shell in read-only mode (only file read, search, and web), so they can't run `git diff`. Paste the diff into the prompt instead.
**Name the project's conventions files.** Read-only Claude castmates run in safe mode and don't load `CLAUDE.md`, and Codex read-only skips the user's `config.toml`. List `CLAUDE.md` / `AGENTS.md` (and any style guide) under Files when conventions matter.
## Delegate (one castmate does a job)
```markdown
# Task
[One paragraph: what to produce, and why.]
## Scope
- Files to read first: [paths]
- Files you may change: [paths or "none — answer only"]
- Out of scope: [what to leave alone]
## Done means
- [Concrete acceptance check, e.g. "rspec spec/models/user_spec.rb passes". Claude castmates can edit files but can't run commands in write mode, so the host runs the check afterward. Codex can run it inside its sandbox.]
- [Output shape, e.g. "reply with a summary of changes and anything you were unsure about"]
## Constraints
- Follow the existing patterns in [path].
- Don't add dependencies, rename public APIs, or reformat unrelated code.
- If something blocks you, stop and say what. Don't guess.
```
## Review (one or more castmates give an independent opinion)
```markdown
# Review request
## Question
[The user's question, in their words.]
## Context
- Files: [paths]
- Recent changes: [one-line summary, then the output of `git diff <base>` in a ```diff block. Keep it under about 50 KB; if it's bigger, include `git diff --stat` plus the hunks that matter.]
- What we already believe: [omit for a blind review]
## Instructions
You are giving an independent review. Be critical and direct; don't hedge.
- Read the referenced files before answering.
- Identify risks, tradeoffs, and blind spots, and suggest a better approach if you see one.
- For each finding, give the file path and function, the evidence, the impact, and the smallest fix.
- Skip style nits unless they hide a bug.
- Use clear headings.
[Paste a preset from presets/ here for a focused hunt.]
```
## Round 2+ addendum (rounds mode)
Append this to the base prompt, followed by the prior outputs inlined under one heading per castmate and round. Keep only the most recent 8 outputs, and cut each to about 1,500 words if needed.
```markdown
## Prior round outputs
Use these to improve quality, not just to avoid duplicates.
- Don't repeat a finding unless you add meaningful new evidence.
- Challenge prior findings: try to invalidate, narrow, or refine the high-impact claims.
- Treat prior findings as leads: follow adjacent code paths, shared utilities, and similar patterns.
- Label any overlapping finding confirmed, refined, invalidated, or duplicate, and say what is new.
### Round 1: codex-sol-high
[output]
### Round 1: fable-high
[output]
```
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |