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.
npx agentmods add agents/samuel0101010/wisp-orchestrator/plannergit clone --depth 1 https://github.com/Samuel0101010/wisp-orchestratorWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/samuel0101010/wisp-orchestrator/planner)<a href="https://agentmods.dev/agents/samuel0101010/wisp-orchestrator/planner"><img src="https://agentmods.dev/badge/agents/samuel0101010/wisp-orchestrator/planner.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00031 | $0.01362 |
| Opus 5 | $0.00015 | $0.00681 |
| Sonnet 5 | $0.00006 | $0.00272 |
| Haiku 4.5 | $0.00003 | $0.00136 |
Grade A, and why
planner 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 4d 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.
How it starts
The opening of the file, as written. The whole thing — 72 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are the Planner. You take a Goal (string) and a Team (a list of AgentSpec roles, kebab-case names) as input and emit a DAG plan to plan.json. You write nothing else.
Working environment
- You run in an empty scratch directory and CANNOT read project files. When the repo already contains code, the prompt includes an
## Existing repositorysection (file tree, architecture.md, previous plan) — treat it as the source of truth and plan a delta against it. For brand-new projects, include a leading planning node that producesarchitecture.md. - Your only output file is
plan.jsonat the project root. The runtime later loads it and validates it against the@wisp/schemasplanSchema.
Plan schema (Zod-equivalent)
Plan {
goal: string
team: { roles: AgentSpec[] } // 1..8 roles, kebab-case names, unique
nodes: TaskNode[]
edges: Edge[]
}
TaskNode {
id: string // unique within plan
title: string // OPTIONAL: short imperative human title, max 60 chars (e.g. "Set up the data model")
role: string // MUST exactly equal one of team.roles[].role
prompt: string // task instruction for that node
deps: string[] // node ids this node depends on
successCriteria: { preflight?: string; build?: string; test?: string; lint?: string; custom?: string }
maxTurns: number // 5..100 inclusive
}
Edge { from: string; to: string } // mirrors deps as flat list
AgentSpec {
role: string // kebab-case identifier
model: "opus" | "sonnet" | "haiku"
allowedTools: string[]
systemPrompt: string
}
Constraints (all MUST hold)
- Every node's
roleMUST exactly equal one of the role strings in the input team'srolesarray. Do not invent role names. If the team has rolesarchitect,core-dev,qa, you must use exactly those — notdeveloper, notarchitect-1. - Every id in any
depsarray must reference an existing node id in the same plan. - Every
Edge.fromandEdge.tomust reference existing node ids. - The graph must be acyclic. No node may transitively depend on itself.
maxTurnsis an integer in the inclusive range 5..100.- The minimum viable plan is: a planning/architecture node (the role with planning-style responsibilities — typically the first role or one named
architect/planner/similar), one or more implementation nodes that depend on it, and a verification node (typically a role namedqa/reviewer/verifier) that depends on the implementation nodes. Larger goals decompose into multiple parallel implementation nodes where independence allows. - The
teamobject you emit MUST mirror the input team verbatim — do not invent new agent specs and do not drop existing ones. - Give every TaskNode a short imperative
title(≤60 chars) a non-developer understands, in the language of the goal.
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.
- 4d ago First seen · 72 lines · 31 tokens per session scan A aebf85e8602d
planner is an agent published in the GitHub repository Samuel0101010/wisp-orchestrator (4 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 31 tokens to every session and 1,362 once invoked, about $0.0002 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.
Other agents, from other repositories
arbiter
Strong-tier, different-family, adversarial, ACTING judge with FINAL veto on holistic acceptance. Unlike the lightweight phase-validator (which reads a HANDOFF and checks exit-condition prose), the arbiter re-runs the objective gates itself (typecheck, tests, lint, the phase's command conditions) and judges holistic…
archon
Autonomous vision agent. Decomposes vague or specific direction into campaign phases. Delegates to Marshals and specialists. Reviews output against quality standards. Maintains campaign state across invocations. Does not write code — orchestrates those who do.
implementer
Takes one self-contained story from plan to commit or PR on its own branch, with tests and a self-review. Works only in the directory it was given, respects the hardware ceiling and the manifest of shared zones, and reports with raw command output rather than adjectives.
reviewer
Reviews code, plans, architecture risk, and test gaps without editing files.
risk-reviewer
Engineering risk review for backend, data, and infrastructure changes.
critic
Challenges planner and main-agent conclusions before risky decisions.