More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Autonomous Project Manager: Long Projects That Don't Lose Their Place

    1

    Keep a multi-day project on track across interrupted sessions: one checkpoint file, clear done-criteria, task history, stuck-loop recovery, and re-checks after every change.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Autonomous Project Manager: Long Projects That Don't Lose Their Place

    Autonomous Project Manager: Long Projects That Don't Lose Their Place

    Example session with this skill installed

    I am building a small inventory app for our community bike workshop over the next four weeks, in evening sessions with an AI helper. I keep losing my place between sessions, I repeated a fix twice already, and I am worried my next change will break the parts that already work. Set up a way to manage this project so I can stop and restart without losing track.

    • Read your context and instructions
    • Compiled the autonomous project manager

    CHECKPOINT — Bike Workshop Inventory App — Samira Haddad — updated after session 5, Nov 12.

    Objective: volunteers can catalogue a bike part, check it out to a borrower, and check it back in, on a phone. Done-criteria: (1) a volunteer can add a part with a photo and shelf location; (2) a borrower can check a part out and the part shows as out; (3) the part can be checked back in and shows as available; (4) the main paths still work after the final change.

    Task list: catalogue — done, checked Nov 8; check-out — in progress; check-in — not started, depends on check-out; search — cut, recorded Nov 10 as a future option, not needed for the objective.

    History: check-out failed twice with the same fix — the loan saved but the part still showed as available. Loop named Nov 11. Recovery: changed approach, traced where the part status is set, found the update went to a copy of the record. Fixed Nov 12; the part now shows as out after checkout.

    Re-checks Nov 12: catalogue still works — added a brake cable, photo and shelf saved. Check-out path re-run end to end and passed.

    Open items: check-in waits on check-out finishing; Samira to confirm the return shelf wording. Next step: build the check-in button so it acts on the checked-out part, then re-run the full path — catalogue, check-out, check-in — before starting anything else. The project is not done: criteria 3 and 4 are still unmet.

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

    About this skill

    Long projects usually fail quietly: a session ends mid-task and the next one cannot tell what was finished, finished work gets repeated, a change breaks a part that worked last week, and the same approach is tried three times because nobody wrote down that it already failed twice. This skill stops that drift with a written record instead of memory. It starts by stating the objective and the observable done-criteria up front, then splits the work into tasks with dependencies and exactly one state each — not started, in progress, done, blocked, or cut. A single checkpoint file carries the objective, the task list with a short history per task, decisions, open items, stuck loops, re-check results, and the exact next step, so any session can stop at any point and the next one can pick up without guessing. The seven-step session protocol covers reading the checkpoint first, stating a one-sentence session goal, working one task at a time while logging as you go, naming a stuck loop after two failed attempts with the same approach, re-checking earlier parts after each change, updating the checkpoint before stopping, and closing out with a handover record. A stuck-loop decision table gives the three recoveries — change approach, narrow scope, or escalate with evidence — and says what to record for each. The skill includes a copy-ready checkpoint template, a task-state table, a complete worked example following a neighbourhood tool-lending app across eight evening sessions, edge cases including stale checkpoints, vague criteria, helper changes, and quiet scope growth, anti-patterns, and an eight-item final check before delivery.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Productivity.

    • It disciplines how a project is managed; it does not do the work, run your checks, or judge whether the objective itself is the right one.
    • It depends on the checkpoint file being updated at the end of each session; a session that skips the update leaves the next session working from memory.
    • It cannot recover work that was never recorded — history only exists for attempts that were logged while they happened.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 3 days ago

    Permissions

    No special permissions declared or detected

    Listed3 days ago

    What's inside

    Frequently Asked Questions

    A to-do list tracks tasks. This tracks the whole project state: the objective, what counts as finished, what was tried and what happened, what is stuck and how it was recovered, what was re-checked after each change, and the exact next step. A to-do list tells you what is left; a checkpoint file tells the next session where it stands.

    That is exactly when it pays off. The longer the gap between sessions, the more the record matters. Read the checkpoint first, work one session goal, update the file before stopping — a restart after weeks away becomes a single reading session instead of a fresh start.

    Yes, in order, but most steps take a minute once the file exists: read it, state the session goal, log as you go, update it before stopping. The steps that take longer — naming a stuck loop, re-checking earlier parts — are the ones that save the most rework.

    New in Productivity