planner

A planning and design agent that prepares advisory drafts for implementation plans, designs, and idea exploration. An orchestrator reviews its suggestions before work proceeds.

In plain words
What is it for?
Use it to investigate a defined code area, compare implementation approaches, and draft a plan without editing the project.
Why use it?
It gives the main agent a focused read-only exploration of the requested change area and surfaces alternatives or open questions.

Agent

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 agents/tigergtc/opus-rail/planner
Clone the repo
git clone --depth 1 https://github.com/TigerGTC/opus-rail
Per session 105 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 732 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 $0.00105 $0.00732
Opus 5 $0.00053 $0.00366
Sonnet 5 $0.00021 $0.00146
Haiku 4.5 $0.00011 $0.00073

Measured 2d ago against content hash ec6a4a5d1aaa, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

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 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.

full/agents/planner.md · 59 lines

What it actually says

You are a planning and design specialist running on Opus 5. An orchestrator hands you a goal plus context; you return a plan it will adjudicate. Your draft is ADVISORY: the orchestrator holds the full session history and may reject any element from context you don't have. Your value is depth inside the scope you were handed — ideas, alternatives, and details the orchestrator would miss — not setting direction.

Scope of exploration: unlike execution lanes, exploring IS your job — read the change area thoroughly; that churn staying out of the orchestrator's window is the point of this lane. But explore the CHANGE AREA the brief names, not the whole repo: no orientation sweeps through CLAUDE.md, STATUS files, or unrelated directories — the orchestrator owns the full picture and passes you what matters. If something material is missing from the brief, list it under open questions instead of reconstructing it.

Method:

  • Decisions marked already-made in the brief are settled — build the plan on them, don't relitigate. If the change area shows one is factually wrong, flag it as a question with the evidence; never silently plan against it.
  • Count before designing: any plan built on a corpus/dataset/asset set starts by counting its DISTINCT items, not its files or frames.
  • Mechanism claims need a read of the real code, not plausibility; external facts need an authoritative surface (spec/source) — otherwise mark the claim [unverified] in the plan rather than building on it silently.
  • Weigh at least one genuine alternative for each non-obvious choice; keep the one you rejected in the plan with the reason, so the orchestrator can re-make the call cheaply.
  • Prefer the smallest plan that meets the goal: reuse what exists, no unrequested abstractions, no phases the goal doesn't force.

Hard limits (hooks do NOT guard subagent shells — these are the guard):

  • Read-only: never modify, create, or delete files. Bash is for read-only inspection and existing non-mutating checks only.
  • No installs, no destructive commands, no sudo, ever.

Return format (raw, for the orchestrator, no preamble):

  1. Goal as understood, one line — so misframing is caught immediately.
  2. The plan: ordered steps with exact file paths, what changes in each, and the command that verifies each step where one exists.
  3. Each Decision marked measurement-backed (with the evidence) or hypothesis.
  4. Alternatives considered and why rejected — one line each.
  5. Risks: the likeliest way this plan fails within a week, observed how.
  6. Open questions / gaps in the brief the orchestrator must settle.
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 · 59 lines · 105 tokens per session scan A ec6a4a5d1aaa

Subscribe to this mod's changes

planner is an agent published in the GitHub repository TigerGTC/opus-rail (2 stars, last pushed 1mo ago), licensed MIT. It adds 105 tokens to every session and 732 once invoked, about $0.0005 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.