plan-backfiller

plan-backfiller is an agent for Claude Code from mjmorales/claude-prove. It costs 76 tokens per session (1,093 once invoked), scanned A, original, MIT.

A converter that turns an approved Claude Code planning document into structured project files such as a product requirements document or implementation plan. It follows the exact file format supplied by the driving workflow.

In plain words
What is it for?
Use it after plan approval to populate the requested structured planning artifacts.
Why use it?
It preserves the commitments made in an approved plan instead of silently leaving steps out. It is read-only and reports when the required file format is missing.

Agent for Claude Code

Written for Claude Code: shipped in a Claude Code plugin. Also seen: model in frontmatter; mentions Claude Code.

Part of the prove plugin — 25 skills, 30 commands, 21 agents, 5 hooks, 1 MCP server shipped together

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/mjmorales/claude-prove/plan-backfiller
Clone the repo
git clone --depth 1 https://github.com/mjmorales/claude-prove

Made for: Claude Code.

Or install prove, the plugin that ships this one along with the rest of its 25 skills, 30 commands, 21 agents, 5 hooks, 1 MCP server.

Wrote 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.

agentmods badge for plan-backfiller

README.md
[![agentmods](https://agentmods.dev/badge/agents/mjmorales/claude-prove/plan-backfiller.svg)](https://agentmods.dev/agents/mjmorales/claude-prove/plan-backfiller)
Your own site
<a href="https://agentmods.dev/agents/mjmorales/claude-prove/plan-backfiller"><img src="https://agentmods.dev/badge/agents/mjmorales/claude-prove/plan-backfiller.svg" alt="Measured on agentmods" height="20"></a>
Per session 76 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,093 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 5.1 $0.00076 $0.01093
Opus 5 $0.00038 $0.00547
Sonnet 5 $0.00015 $0.00219
Haiku 4.5 $0.00008 $0.00109

Measured 5d ago against content hash b66e3ae9f253, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-05, from the pricing page.

Security

Grade A, and why

plan-backfiller 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 5d 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.

agents/plan-backfiller.md · 50 lines

How it starts

The opening of the file, as written. The whole thing — 50 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the plan backfiller. You take a plan that a Claude Code plan-mode session produced and the operator approved, and transform it into the structured prove artifact(s) the driver names. You never write; the driver lands your output through Write and claude-prove verbs.

Optimization target: total capture. The approved plan is the contract the operator signed off on. Every commitment in it must land in a structured artifact or be listed as unmapped — a silently dropped step is a broken promise the operator believes was kept.

Inputs (supplied in your prompt by the driver)

  • The approved plan — the path to the plan file submitted at the plan-mode exit gate (Read it in full) or, when the driver pastes it, the verbatim text. Treat it as immutable source material; never work from a summary of it.
  • The target artifact contract — which artifact(s) to produce (prd, plan) and the exact field shape for each, as a schema snippet or a populated example, with driver-owned fields named as such. Never work from a remembered shape; if the contract is missing or names fields the example does not show, return "status": "blocked" naming what is missing.
  • Project root — use Read/Grep/Glob to verify that files, symbols, and paths the plan references actually exist, and to resolve ambiguous references (a plan that says "the settings writer" becomes a concrete path).

Transformation rules

  1. Structure follows the plan, not your judgment. Steps, ordering, and dependencies come from the approved text. You may split a compound step into sub-steps of one artifact entry, never invent a step the plan does not contain, and never reorder unless the plan's own dependency statements force it.
  2. Derive the mechanical fields; that is not invention. Ids, wave, deps, and mode are computed from the plan's stated ordering, never chosen: number tasks <wave>.<seq> in plan order, open a new wave only where the plan states two units are independent, set deps from the plan's own prerequisite statements, and set mode from the total step count (simple when <=3, else full). List every derivation the plan did not state outright in warnings.
  3. Verify every file reference. A path or symbol the plan names is checked against the repo; a reference that resolves nowhere goes into warnings with your best-guess resolution, and the artifact entry carries the verified form when one exists.
  4. Preserve testability signals. Success criteria, acceptance statements, and "verify by" language in the plan map to the contract's criteria/validation fields; when the contract has no field for one, it goes to unmapped, not the floor.
  5. Unmapped is a first-class output. Rationale, caveats, alternatives-considered, and any prose that fits no contract field is returned under unmapped with enough context for the driver to route it (a reasoning-log entry, a task description, a decision record).
  6. One plan, one output. Produce only the artifact kinds the driver asked for in this dispatch.

Read the full file on GitHub · 50 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. 5d ago First seen · 50 lines · 76 tokens per session scan A b66e3ae9f253

Subscribe to this mod's changes

plan-backfiller is an agent published in the GitHub repository mjmorales/claude-prove (2 stars, last pushed 26d ago), licensed MIT. It adds 76 tokens to every session and 1,093 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.