CocoClaw Context Relay — Resume Work Across Agent Sessions

    1

    Save a local project checkpoint with goals, decisions, verification evidence, unfinished work, and the next action. Verify it against the current files when a new agent session resumes.

    Free

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    CocoClaw Context Relay — Resume Work Across Agent Sessions

    Example session with this skill installed

    Resume the saved task in this project using CocoClaw Context Relay. Verify the checkpoint against the current files, continue the next action, keep changes uncommitted, and preserve existing files.

    • Read your context and instructions
    • Compiled the cococlaw context relay

    Fresh-session recovery result

    Local acceptance demonstration with project-relative file references. This is the complete outcome of the resume run; separate stale-state tests were performed afterward.

    Resumed from .context-relay/checkpoints/20261029T090412Z-b6fad2a9.md.

    • Verified the project root, branch relay-acceptance, HEAD 2c3227e485bfc17c6b1ed2fe3563afc2705e2153, and all five recorded file digests.
    • Recovered the objective: remove surrounding whitespace from greeting names while preserving ordinary names.
    • Confirmed the code change had already been completed and README documentation was the remaining action.
    • Updated README.md with the two observed input/output examples below.
    • Reran python3 -m unittest discover -s tests -v: two tests passed, exit 0. git diff --check also passed.
    • Preserved the source edits, pre-existing untracked notes, and checkpoint. Nothing was staged or committed.

    Complete README delivered by the resumed session

    
    # Greeting utility
    
    A small utility that greets a supplied name.
    
    `greeting(name)` removes surrounding whitespace from the supplied name. Ordinary
    names remain unchanged:
    
    ```python
    greeting("  Ada  ")  # "Hello, Ada!"
    greeting("Lin")      # "Hello, Lin!"
    ```
    

    Verification limits

    The result covers the existing two greeting tests and the recorded file contents at the time of the resume run. It does not claim that later edits or additional edge cases were tested. The workflow used explicit SKILL.md loading in independent Codex agent contexts; other clients and automatic skill discovery were not tested.

    edge cases were tested. The workflow used explicit SKILL.md

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

    About this skill

    Long-running coding tasks often lose context when a session restarts. CocoClaw Context Relay gives your agent a focused workflow for saving what matters and checking it before continuing.

    Create a local checkpoint containing the project identity, objective, decisions, working changes, observed test results, blockers, and next action. In a later session, the agent checks the project and current files before relying on the saved state. Earlier test results remain tied to the contents that were checked.

    Includes:

    • Checkpoint, resume, and handoff instructions.
    • A practical checkpoint format and evidence rules.
    • Acceptance scenarios for stale state, a wrong project, and interrupted checkpoints.
    • Fresh checkpoint files that preserve earlier saved state.

    Requirements: an agent with local file access that can load SKILL.md instructions. Git is optional and provides additional repository evidence. No CocoClaw server, Docker container, external account, or API key is required.

    This is an instruction workflow. It does not run as a background memory service or synchronize sessions. A new session needs access to the saved checkpoint. Recovery depends on the checkpoint and evidence available in the project.

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

    Requires access to a usable local checkpoint and the selected project files. Does not synchronize sessions or run automatically in the background. A newer incomplete checkpoint can be skipped in favor of an earlier complete one; missing evidence must be recovered or reported. Earlier test results do not validate later file edits. Automatic skill discovery, non-Git behavior, and clients outside the tested Codex contexts have not been validated.

    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

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 19 days ago

    • Free with an Agensi account

    Nothing to install. It runs in your AI agent.

    Optional
    GitTerminal accessFile hashing tool

    Git adds repository evidence; the shell and hashing tool make content fingerprints.

    Permissions

    Terminal / Shell
    Read Files
    Write Files

    File Scopes

    **/*

    Reads task-relevant files in the project selected by the user. Creates local checkpoint documents under .context-relay/checkpoints/ or a user-selected location. Uses Git/status and content hashes when available. No required external accounts, network access, environment-variable access, hooks, or background services. Resume only changes project files as part of the user's existing task.

    Listed19 days ago

    What's inside

    Frequently Asked Questions

    The objective, decisions, working state, observed verification evidence, unfinished work, blockers, and next action. A new session reads the checkpoint and verifies it against current files before continuing. Recovery depends on the saved information and evidence available in the project.

    Explicit SKILL.md loading was tested in independent Codex agent contexts for save/resume, stale working contents, wrong-project refusal, and incomplete-checkpoint recovery. Other clients and native automatic discovery have not been tested. The host agent needs local file access.

    It instructs the agent to compare the resolved project directory, repository identity, branch, HEAD, and current files before continuing. The tested wrong-project scenario stopped before changes. This is an instruction workflow, so it does not provide a runtime-enforced guarantee.

    The agent compares current contents with the saved evidence, including file digests when available, and inspects changes. Earlier test results remain tied to the contents that were actually tested. A matching commit alone does not validate newer uncommitted edits.

    Checkpoints are local files, by default under .context-relay/checkpoints/ in the selected project. This workflow does not independently upload project files. The host agent, LLM provider, and marketplace delivery service may use cloud connections under their own policies; local checkpoint storage does not mean the entire agent runs locally.

    SKILL.md instructions for checkpointing and resuming, a checkpoint format reference, acceptance scenarios, and a short README. No CocoClaw server, Docker deployment, external account, or API key is required by the workflow itself.

    New in Agents & Orchestration

    Free