agentsclimarketplace

Planf3

Skill tony-adamson/groundwork/codex/skills/planf3

Evidence-first process skills for coding agents: codebase analysis, solution design, minimal implementation plans

Install
npx -y skills add tony-adamson/groundwork --skill planf3

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Use when explicitly asked for $planf3 to create, update, or build a minimal Markdown implementation plan from SOLUTION.md or a concrete approved task. Do not use for solution design or speculative architecture.

SKILL.md

3.6 KB, 778 tokens by cl100k_base, as published. Nobody here has run it

PlanF3 → minimal implementation plan

PlanF3 works in two modes:

  1. Create Plan — create specs/<descriptive-name>-implementation-plan.md.
  2. Build Plan — execute an already-approved plan with the READY_FOR_BUILD status.

The plan is always Markdown-first. Do not create presentation or visual artifacts. If a diagram is needed — use short text, or Mermaid only when it genuinely helps.

Primary goal

PlanF3 does not search for "the best architecture". It compiles the approved SOLUTION.md into the smallest executable plan that implements the requirements, preserves the contracts, and has reproducible checks.

If SOLUTION.md exists, it outranks the plan. Do not reinvent the architecture.

Output file

Defaults:

  • PLAN_OUTPUT_DIRECTORY: specs/
  • PLAN_FILE: specs/<descriptive-kebab-name>-implementation-plan.md

The plan is two-block:

  1. Block 1. For the human — a short readable summary.
  2. Block 2. For the agent — a detailed execution contract.

Statuses

  • DRAFT
  • BLOCKED_FOR_SOLUTION_AMENDMENT
  • BLOCKED_BY_SCOPE_OVERDESIGN
  • READY_FOR_BUILD
  • BUILD_IN_PROGRESS
  • BUILD_COMPLETE
  • BUILD_FAILED

Minimal Diff Compiler Mode

When SOLUTION.md exists:

  • do not improve the architecture;
  • do not add future-proofing;
  • do not add a dependency/subsystem/state/abstraction outside SOLUTION.md;
  • do not expand the scope;
  • do not turn a local feature into a platform;
  • do not add more docs/tests than needed to prove the behavior;
  • if the plan needs a new architectural decision — BLOCKED_FOR_SOLUTION_AMENDMENT.

Modes

If the request contains a path to an existing plan and asks to implement it — read build-plan.md.

If the request asks to create a plan from SOLUTION.md or a task — read create-plan.md.

If the request asks to update a plan — read update-plan.md.

If the request asks to update references between plans — read update-references.md.

Before READY_FOR_BUILD, always run plan-readiness-gate.md.

Mandatory gates before READY_FOR_BUILD

  • correctness gate PASS;
  • executable validation gate PASS;
  • minimality gate PASS;
  • no architecture decision deferred to the builder;
  • no speculative architecture beyond SOLUTION.md;
  • no material requirement existing only in prose;
  • no validation based only on static text when behavior can be checked.

Build Plan

During implementation:

  • execute the phases sequentially;
  • make the minimal diff;
  • use subagents adaptively;
  • after non-trivial phases, run a fresh read-only verifier;
  • fix only confirmed blocking findings;
  • stop on a new design decision, scope overdesign, or an unsafe action.

Harness adaptation

If the harness does not provide an isolated-subagent tool (for example, Pi) — perform the implementer/verifier/challenger roles inline as separate passes: the verifier pass outputs only findings, and the coordinator responds. Do not simulate spawning subagents and do not claim they were launched.

Language

Write plans and reports in the user's language — the language of the user's request and conversation, not the language of these instructions. If the request is in Russian, the plan is in Russian. Do not translate file names, commands, APIs, symbols, statuses, or code identifiers.

What ships with it: 6 files

12.0 KB alongside SKILL.md

agents/

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.