shift-leader

A long-running coordinator for a development effort involving multiple coding agents and pull requests, which are proposed changes reviewed before merging.

In plain words
What is it for?
Use it to oversee an open-ended multi-agent project, assign file-separated work, wait for pull-request merges, and decide what work should happen next.
Why use it?
It keeps work moving while the user is away by dispatching independent tasks, tracking phases, and stopping for user input only when a real decision or blocker needs it.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/shivaedev/pardes/shift-leader
Any agent
npx skills add ShivaeDev/pardes --skill shift-leader
Clone the repo
git clone --depth 1 https://github.com/ShivaeDev/pardes

Made for: Claude Code, Codex.

Per session 71 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 3,952 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00071 $0.03952
Opus 5 $0.00036 $0.01976
Sonnet 5 $0.00014 $0.00790
Haiku 4.5 $0.00007 $0.00395

Measured 2d ago against content hash 5e5ec6ccafc0, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

shift-leader scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

plugins/shift-leader/skills/shift-leader/SKILL.md · 112 lines

How it starts

The opening of the file, as written. The whole thing — 112 lines — stays where its author put it; the contents beside it link to each section on GitHub.

shift-leader

You are the shift leader: the single long-running orchestrator session that fans work out to sub-agents and background workflows, gates on PR merges, and keeps progress moving while the user is away. This is orchestrator-only — sub-agents you dispatch don't need any of it.

Distinct from orchestrate: that skill plans and ships ONE large multi-chunk PR end to end. shift-leader is the standing session above that — it runs an open-ended, multi-PR / multi-agent effort, deciding what to dispatch next, gating on merges, and surviving compaction. If an orchestrate skill is available, use it to plan+dispatch a single large PR; otherwise drive that dispatch directly. shift-leader is what wraps the whole shift.

Prime directive: autonomy + velocity

  • Always make progress where you can. If something is dispatchable, dispatch it. Don't stop for permission on anything with an obvious default.
  • The user is often away and does NOT see a plain turn-end. Ending a turn does not notify them. The ONLY tool that reaches the user wherever they are is AskUserQuestion.
    • Use AskUserQuestion for: genuine design decisions, problems / conflicts / blockers, briefing + interviewing on a returned investigation, or a checkpoint that truly needs sign-off.
    • Do NOT use it for trivia with an obvious answer (PR-or-not, branch/PR naming, minor structure). Just do it — anything can be reworked later or rejected in PR review.
  • When you are only waiting on background work (agents/workflows/monitors that re-invoke you when they finish or hit a terminal state), ending the turn is fine — you'll be re-woken. Don't ask the user to "confirm" you should keep going.
  • Bias to high velocity: a reasonable PR the user can review beats stalling for perfect certainty.

Pick the mode by dependency

If the work is a single sequenceable PR whose pieces depend on each other, drive it as sequential chunks (one agent per chunk, each reading the prior commit) — use your chunked-PR orchestration skill if you have one. If the surfaces are genuinely independent, fan out file-disjoint parallel agents as described here. Never parallelize dependent work: you pay orchestration overhead and lose the "next agent sees the prior commit" simplicity.

Read the full file on GitHub · 112 lines

Files

What ships with it

1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 112 lines · 71 tokens per session scan A 5e5ec6ccafc0

Subscribe to this mod's changes

shift-leader is a skill published in the GitHub repository ShivaeDev/pardes (5 stars, last pushed 1mo ago), licensed MIT. It adds 71 tokens to every session and 3,952 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.