leek/agent-skills / skills/panel
panel
VerifiedConvene a panel of agent CLIs (grok, claude, codex, agy) on a prompt or a code review, then report one consensus grouped by how much they agree.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 7.3 KB |
| agents/openai.yaml | config | 245 B |
| references/cli-invocations.md | markdown | 4.9 KB |
| references/panelist-brief.md | markdown | 2.0 KB |
| references/review-shape.md | markdown | 2.0 KB |
| scripts/launch.shexecutable | script | 1.8 KB |
references/panelist-brief.md
View on GitHub# Panelist brief
The brief every panelist subagent gets in [`SKILL.md`](../SKILL.md) step 3, beside the
path of the **same** task file. Give it in full on every run.
- **Launch with the launcher script, literal paths only.** Run your CLI **once**, from
the repo root (or your worktree), with `bash LAUNCHER <cli> TASK_FILE SCRATCH/panel-<cli>.out`,
where `LAUNCHER` is this skill's [`scripts/launch.sh`](../scripts/launch.sh), and every path is
absolute and written out in full: worktree isolation refuses a command whose words are
computed at runtime (`$(cat …)`, `$TASK`, `$PWD`). The script holds each CLI's headless
command and flags (no `--model`, ever) and writes all its output to the `.out` file. If a
guard still refuses the launch, that is a launch-form problem, not a CLI error: say
which word it refused; `bash LAUNCHER --print <cli> …` prints the plain command to run instead.
- **Run it detached: never let one time out.** These CLIs take many minutes, past the
shell tool's 10-minute foreground ceiling, which kills a run before it reports. Launch
with the Bash tool's `run_in_background: true`; your turn pauses and you are re-invoked
when the CLI exits; then read the `.out` file. Where the shell tool has no background
mode and rejects `&` (Grok CLI), run it in the foreground with the tool's timeout set to
its maximum (for example `timeout: 900000`), never its default.
- **Wait with one call, never poll.** No `sleep`, `ps`, `pgrep`, or repeated reads of
the output file while the CLI runs: each check spends a turn, and the turn cap ends the
subagent while its CLI is still running.
- Distill the output into a short, normalized result. For each point the CLI makes,
capture three things: a one-line **claim**, a **location** (`file:line`, or
`general`), and a one-line **reason**. Drop the CLI's logs, thinking, and prose.
- Return `{ cli, status, points[] }`. `status` is `responded`, or `errored` with a
one-line cause (a non-zero exit or an empty `.out`). Return the object only, no narration.
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 |