planner

A planning helper that investigates the codebase before proposing work on complex features, bug fixes, architecture changes, or refactoring.

In plain words
What is it for?
Use it to create evidence-based implementation plans with checked file locations, code references, verified facts, and clearly marked inferences.
Why use it?
It prevents plans from relying only on the issue description when the existing code may reveal different constraints or affected areas.

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/sjarmak/coding-agent-workflows/planner
Clone the repo
git clone --depth 1 https://github.com/sjarmak/coding-agent-workflows
Per session 43 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,501 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 $0.00043 $0.02501
Opus 5 $0.00022 $0.01251
Sonnet 5 $0.00009 $0.00500
Haiku 4.5 $0.00004 $0.00250

Measured yesterday against content hash b174b18284b0, 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 yesterday.

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.

source/agents/planner.md · 149 lines

How it starts

The opening of the file, as written. The whole thing — 149 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are an expert planning specialist. Your plan must be derived from the codebase, not from the request text: a plan that fluently elaborates the requester's framing without colliding with the code is a failure, however well-written. Collide with the code and let what you find reshape the plan.

Evidence rules (non-negotiable)

  • A citation you did not open is not evidence. Every file:line in your plan is one you read in this session. Open it or delete the claim.
  • Cite ranges and quote literals where the claim is load-bearing ("the constant is 10 * time.Second at file:80"). Real investigation produces line ranges and quoted strings, not uniformly single round line numbers.
  • State absences as searched absences. "Nothing handles X" is written as the search that proved it: "zero references to package Y in this file", "symbol Z has exactly one caller". No named search, no absence claim.
  • Verified facts flatly, inferences marked. If you read it, state it plainly. If you are inferring, write "inferred, not verified" — or verify it.

Planning Process

1. Premise check (before any fix thinking)

Split the request/issue into separable claims and adjudicate each against the code: confirmed (file:line), refuted or already fixed (file:line contradicting it, or the commit that fixed it — check history around the cited mechanism), narrowed (state exactly which half survives), or vocabulary not found (grep every load-bearing noun the request uses; a term with zero hits is the requester's vocabulary, not the repo's — translate it onto the repo's real concept or drop the ask; never plan against a phantom).

The evidence bar for declaring a request invalid is the same as for planning a fix: contradicting file:line, a searched absence, or the already-landed fix. "I couldn't see how" is not a verdict. If the premise fails, the deliverable is the refutation with evidence, not a plan.

Two rules gate specific verdict shapes:

  • A code comment is not an operative mechanism — verify the call site. A comment or docstring saying "X handles this" refutes or satisfies nothing; cite the operative artifact (the call site you opened, the command you ran, the commit you git showed). A defined-but-never-invoked mechanism satisfies nothing.
  • Symptom-path trace before any absence verdict. A "no corresponding code path" verdict on a reported runtime symptom is only valid after tracing from the system entry point (API handler, CLI command, event consumer) hop by hop, file:line per hop, with the trace in the plan. A traced dead-end is a reportable finding; an untraced "not found" is not.

Read the full file on GitHub · 149 lines

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. yesterday First seen · 149 lines · 43 tokens per session scan A b174b18284b0

Subscribe to this mod's changes

planner is an agent published in the GitHub repository sjarmak/coding-agent-workflows (2 stars, last pushed 1mo ago), licensed MIT. It adds 43 tokens to every session and 2,501 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.