Cdd Plan logo

Cdd Plan

Community
Oscaner
cdd-plan

Independent plan-writing orchestrator -- Node-anchored flow with digraph as single control-flow source of truth. Consumes the /superpowers:writing-plans(pi:/skill:writing-plans) flow inline as this session's baseline, backfills the design spec on substantive drift before authoring, commits the authored plan before the review-fix loop, and runs the loop following the engine's next face — the engine's next directive carries the pipeline continuation. Callable standalone; invoke as kairos:cdd-plan(pi:/skill:cdd-plan).

Overview

PublisherOscaner
Repositoryskills
Skill namecdd-plan
Stars
63
Forks
0
Bundled files
Instructions only
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by Oscaner on GitHub. Read the source before you install it.

Installation

Install the Cdd Plan AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/Oscaner/skills/tree/main/packages/kairos/skills/cdd-plan
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/Oscaner/skills.git /tmp/skills
mkdir -p .claude/skills
cp -r /tmp/skills/packages/kairos/skills/cdd-plan .claude/skills/cdd-plan
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Cdd Plan in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Cdd Plan on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Cdd Plan is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Kairos CDD-Plan

Writes a plan document from an approved spec, backfills the design when planning surfaces substantive drift, commits the plan before reviewing under the cdd plan contract; on closure the engine's next: face carries the pipeline continuation.

Invocation discipline — every engine command runs as a background dispatch: launch it in the background and await the completion notification — never poll — then read the full output (stdout/stderr). Output filtering is forbidden — no piping to tail/head, no 2>&1 |, no EXIT=$? capture. Round rhythm — one review pass per dispatch; a fix round re-enters the review while the route is not done; commit the authored plan before the first review round; ensure the working tree is clean before entering review (dirty → BLOCKED; the orchestrator writes no tree during dispatch). Failure face — cross-node failure handling lives in the Failure Modes table below (the single source); node Fail entries keep only local behavior.

Flow Digraph

Node Definitions

run-cdd-plan-session

  • Do: Import /superpowers:writing-plans(pi:/skill:writing-plans) to plan the approved spec; it lands the draft plan (session-call; the upstream document is not read)
  • Read: nothing before the import; the import plans from the approved spec and lands the draft plan
  • Exit: Import landed → backfill-design
  • Fail: Upstream superpowers plugin missing → BLOCKED (no downgrade, no skip, no inline restatement)

backfill-design

  • Do: Check the drafted plan against the approved design. Substantive drift (factual error / missing constraint / a new implementation step the design does not cover) → backfill the design spec first — revise the spec + record the drift — so spec and plan agree before plan-review (the design-backfill rule). Cross-phase matters still backfill to the parent overall per Boundary rules
  • Read: approved spec + landed plan draft
  • Exit: spec ↔ plan consistent → author-plan
  • Fail: Entering plan-review with an un-backfilled drift → violates the design-backfill rule

author-plan

  • Do: Write the complete plan document to docs/kairos/plans/YYYY-MM-DD-<feature>.md — generate the canonical skeleton first (npx -y @oscaner-skills/cdd-engine@latest schema get plan --output skeleton <target path>) and fill its slots in place; the structure stays as generated. Plan header MUST carry the document identity in its front-matter slug: key, inherited from the approved spec it plans — the same value the spec carries, so spec and plan resolve as one run (the generated skeleton emits the slug: key for you to fill). The plan declares no attribution key of its own: its program chain is implied by the identity document of the same slug, so cdd-report resolves program attribution through that chain (workspace plan record → the plan's slug: → the identity document of that slug, read through its own attribution — a phase-spec's program:, or, for an overall / single-spec, its own inventory), the plan record being the first hop of its program chain.
    • Constraints — the ## Constraints section declares the constraint surface the implement child reads — zero separately materialized constraints file; declaring no constraint rows never blocks a dispatch — no silent fallback, no phantom rules file.
    • Dependencies — declare each task's dependencies as the single directed edge on its task record — - **DependsOn**: <n, n, …> (any existing task id; forward references are legal; none/empty = the explicit no-dependency declaration); the declared edges group the tasks into dispatch waves.
    • Write authority — cross-task adjudication surfaced later (user adjudication / phase-level re-scope) is handled by the orchestrator under Plan write authority (I3): it amends the target task's - **Objective**: / - **Steps**: / - **Acceptance**: data fields directly.
  • Read: approved spec + backfill-design output + the plan's structural contract (the same structure fact the review reads)
  • Exit: File written → self-review
  • Fail: Schema read fails or the draft plan is unusable → BLOCKED (missing plan input)

self-review

  • Do: Review the just-written plan before committing — run five checks, substance first: (1) coverage — every approved-spec requirement maps to a task; (2) step decidability — each step lets the implementer write exactly one reasonable thing (a line that decides nothing, or a signature-and-tests-determined transcript, is a defect); (3) consistency — task ids / DependsOn edges / naming agree; (4) failure-mode coverage — every failure mode the spec implies has a task whose checkable exercises it; (5) proportion — the plan is not many times the length of the spec it implements. Then the mechanical tail: zero leftover :::fill markers, zero angle-bracket literals outside link-anchor slots (cheap hygiene the doc-contract gate would otherwise flag). Issues found are fixed inline on the same file; this is a pre-commit self-check inside this flow, not a review round (no next: dispatch, no findings carrier).
  • Read: the written plan document + the approved spec it implements
  • Exit: Self-check clean (issues fixed inline) → commit-plan
  • Fail: Unresolvable defect needing design input → BLOCKED (missing plan input)

commit-plan

  • Do: git add the plan + conventional commit.
  • Read: The committed plan file path
  • Exit: Commit complete → NEXT-LOOP
  • Fail: Git error → report + fail-open (do not block user plan review)

NEXT-LOOP

  • Do: Run the review-fix rhythm for the authored plan — one review per pass. Dispatch the review round on the current ref (npx -y @oscaner-skills/cdd-engine@latest review --type plan --plan <path>); dispatch the output's next: line as written — the literal is the route: a fix dispatch consumes the findings handoff, a closure ends the flow (follow the engine's next: literal — for a plan it routes the next pipeline entrypoint); a BLOCKED/TIMEOUT round carries no next: line — its reason rides stderr for reporting.
  • Read: the review output — its conclusion and the next: line, plus the findings handoff path
  • Exit: the next: line routes the next pipeline entrypoint (a skill ref / the pipeline terminus) → the flow ends (the engine's next: literal carries the next pipeline entrypoint when one exists); each fix pass re-enters this hub while the route is not terminal — the engine's next: literal has not yet handed the orchestrator off (the until next is terminal self-loop)
  • Fail: Re-running a review after a closure conclusion → violates Review Convergence (stop + report)

Invariants

#Invariant
I1Review Convergence (next: is the dispatch) — a review closes by its conclusion (APPROVED / CHANGES_REQUESTED / REVIEW_FIX), and the output's next: line carries the engine's default next step as a complete executable command; read the next: line and dispatch it as written — the literal is the dispatch, no kind→command mapping layer. A mid-backfill or a user adjudication that lands governs over the suggestion (current world state wins). Fixes always dispatch via the fix round, consuming the findings handoff; the orchestrator must not edit in place as a substitute; a fix commit opens a new review, never a re-run of a closed one
I2Plan commit discipline — the authored plan commits before the first review round; each fix round commits its own changes — approval closes the loop on a committed state; do not wait for dev merge
I3Plan write authority — the plan's writes discriminate by trigger: review findings land through the fix round dispatch only; a user adjudication or durable backfill lands through the orchestrator's own standalone commit, amending the target task's data fields directly; fix/implement agents hold zero plan-modification authority
I4Mid-Flight Backfill — a user-raised backfill of overall/spec/plan docs surfaced while a dispatch is in flight lands through four ordered steps on the current call's return: (1) land immediately — hot context, no deferral; (2) commit on its own — a standalone change, never mixed with implementation commits; (3) pause the loop until clean — an uncommitted backfill trips the next review's entry gate (dirty → BLOCKED); (4) audit in-band on resume — the next review audits the backfill in-band

Failure Modes

failurebehaviorreason
Upstream superpowers plugin missingBLOCKED (install superpowers — see the kairos README's installation table)Block policy: no silent fallback
review re-run after a closure conclusion (REVIEW_FIX / APPROVED)Violates I1 (Review Convergence) — stop + report to userA fix commit opens a new review, never a re-run of a closed one
Entering review with an un-backfilled driftViolates the design-backfill rule — stop + backfill firstSpec and plan must agree before review
Git commit errorreport + fail-openDo not block user plan review
Completed round with no next: line (not BLOCKED/TIMEOUT)HARD_ERROR — report the carried reason, re-run the same command to continueNo dispatchable next

Frequently asked questions

What does the Cdd Plan AI skill do?

Independent plan-writing orchestrator -- Node-anchored flow with digraph as single control-flow source of truth. Consumes the /superpowers:writing-plans(pi:/skill:writing-plans) flow inline as this session's baseline, backfills the design spec on substantive drift before authoring, commits the authored plan before the review-fix loop, and runs the loop following the engine's next face — the engine's next directive carries the pipeline continuation. Callable standalone; invoke as kairos:cdd-plan(pi:/skill:cdd-plan).

Why use Cdd Plan on TypingMind?

Because you install it once and use it with any model. Cdd Plan is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Cdd Plan in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/Oscaner/skills/tree/main/packages/kairos/skills/cdd-plan. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Cdd Plan?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Cdd Plan?

As many as you like. As long as a model supports skills, you can use Cdd Plan with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Cdd Plan AI skill free?

Yes. It is published on GitHub by Oscaner under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇