More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+22 more

    İskele — Project Planning Kit: Roadmap, Backlog, Tracker

    1

    Turns a vague project idea into a gated roadmap, an atomic backlog with testable acceptance criteria, an Excel tracker and a progress report computed from it.

    Free

    Security scanned
    İskele — Project Planning Kit: Roadmap, Backlog, Tracker

    İskele — Project Planning Kit: Roadmap, Backlog, Tracker

    Example session with this skill installed

    While a process is being executed, can another user step into the same task? And how does it work when the work moves between two or three departments?

    • Read your context and instructions
    • Compiled the skele
    • Generated the document

    Translated excerpt from a real project's backlog, built with İskele (Turkish original, shortened).

    F8 — Multi-department execution and work pool

    Born from the question above. Reading the code found two gaps:
    R14 — no lock during execution: two people completing the same step silently overwrites the first person's signature.
    R15 — a process belongs to one department; cross-department work is neither supported nor explicitly ruled out.
    Order is mandatory: F8.1 → F8.2 → F8.3. F8.1 stops today's data loss on its own.

    Epic F8.1 — Concurrency and signature integrity (R14)

    • F8-BE-01 (S) Close the same-state transition on terminal states.
      • Accept: User A marks a step done; User B requests the same transition → 409 already_resolved, the response says who resolved it and when, and DoneBy is still User A.
    • F8-BE-02 (M) Optimistic lock (row_version) and 409 stale_record.
      • Accept: two concurrent requests → one 200, one 409. A request without a version is refused — accepted silently, the protection would be optional.
    • F8-FE-01 (S) One conflict surface. Depends: F8-BE-02
      • Accept: a 409 shows a state, not an error — "User A completed this step at 14:32" + Refresh — and the note the user was typing is not lost.
    • F8-X-01 (S) Concurrency test suite — the arbiter.
      • Accept: double completion, stale-version writes and writes to a closed instance are tested. Until now this class had zero tests, which is why R14 went unnoticed.

    And then: the tracker and the progress report follow the backlog. In this project the report reads 184 of 186 tasks done, 97% effort-weighted, computed by progress.py from tracker.xlsx — see the screenshots.

    tracker.xlsx

    PDF · document

    Generated

    Example file from a real run - the skill writes it into your workspace.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    You describe what you want to build. İskele gives you back a plan you can execute and track — not a page of prose.

    Most AI planning output fails in one of two ways: a handsome roadmap where "what do I do today" is unclear, or a task list whose progress is estimated by feel. İskele closes both.

    What you get

    A twelve-piece kit, written to files in your project:

    • Domain model — it looks for the two things in your domain that get conflated but have separate life cycles, before any schema is written.
    • Phased roadmap with gates — phases are a dependency chain, each with an explicit out-of-scope list and a go/no-go gate.
    • Atomic backlog — every task has an ID, an S/M/L estimate, its dependencies and an executable acceptance criterion ("an unauthorised request gets 403", never "works well"), written from the user's side.
    • Quality gates — Definition of Done, phase go/no-go, and recovery ramps for when a task will not close.
    • tracker.xlsx generated from the backlog, and an HTML progress report whose numbers are computed from the tracker (effort-weighted), not typed by hand — with labels in English or Turkish.
    • ADR log with a checker, and an AGENTS.md signpost for the next agent that opens the project.

    Why it is different

    • Asks at most three questions, and every question comes with its own recommendation.
    • Refuses false precision: estimates are labelled as estimates, with their basis.
    • Every acceptance criterion is a refutation condition written before the work — so the backlog hands off cleanly to evidence-based review (sibling skill: Mizan).

    Use it when

    "Plan this project", "give me a roadmap / backlog", "I don't know where to start", "MVP scope", "set up a progress tracker". Not for a single-file change or an urgent bug — just do those.

    Requirements

    Works with any agent that reads SKILL.md folders. The tracker and report scripts need Python 3 and openpyxl; without them the skill says so and offers a CSV tracker instead of pretending.

    Open source (MIT, docs CC-BY 4.0). Docs and examples: https://xinmurat.github.io/Iskele/

    Security notes

    The scripts are local file tools (markdown → xlsx → html). The only subprocess call is git rev-parse --short HEAD in scripts/progress.py, used to stamp the report with the commit it was generated from. No network access, no environment variables, no obfuscated code.

    How to use

    Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill passes our trust checks below.

    Security scanned

    Passed our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free forever

    No account required to browse

    Trust & safety

    Security scanned

    Verified clean 10 days ago

    • Free to download with an account

    Listed10 days ago

    What's inside

    Frequently Asked Questions