Agent-Skills.md

Agent Skills: Chain System

Create, list, and load multi-session chain links that capture work context for later continuation. Use when you want to save current progress, resume a project from a prior session, or inspect chain history. Triggers: chain link, chain load, chain list, save context, resume work, continue later, session handoff.

UncategorizedID: DeevsDeevs/agent-system/chain-system

Repository

DeevsDeevs/agent-system
DeevsDeevsLicense: Apache-2.0
205

Install this agent skill to your local

pnpm dlx add-skill https://github.com/DeevsDeevs/agent-system/tree/HEAD/chain-system

Skill Files

Browse the full folder contents for chain-system.

Download Skill

Loading file tree…

chain-system/SKILL.md

Skill Metadata

Name
chain-system
Description
"Create, list, and load multi-session chain links that capture work context for later continuation. Use when you want to save current progress, resume a project from a prior session, or inspect chain history. Triggers: chain link, chain load, chain list, save context, resume work, continue later, session handoff."

Chain System

Overview

Capture durable, structured summaries of work so a later session can resume without re-discovery. Provide three subcommands under one skill: link, load, and list.

Usage

Invoke with a subcommand:

$chain-system link <chain-name>
$chain-system load <chain-name>
$chain-system list

If the chain name is missing for link or load, ask the user for it.

Chain Root

Default to .claude/chains to remain compatible with the existing Claude chain-system plugin. If the user explicitly wants Codex-only storage, switch to .codex/chains and confirm.

Subcommands

link

Create a new chain link summarizing the current work.

Steps:

  1. Build a concise but complete summary using this structure:
    • Primary Request and Intent
    • Key Technical Concepts
    • Work Completed
    • Decisions and Rationale
    • Files and Code Changes (Created/Modified/Read)
    • Unresolved Issues and Blockers (only if any)
    • Pending Tasks
    • Current Work
    • Next Step
  2. Create a slug (kebab-case) and a human readable title.
  3. Create the directory: mkdir -p .claude/chains/<chain-name>.
  4. Generate timestamp in local time: date '+%Y-%m-%d-%H%M'.
  5. Write the summary to: .claude/chains/<chain-name>/<timestamp>-<slug>.md
  6. Write only the summary (no analysis), then tell the user the file path and how to load it.

Use the same file format each time so load can extract summary and next step reliably.

load

Load the most recent chain link for a named chain.

Steps:

  1. Find the most recent link:
    • ls .claude/chains/<chain-name>/*.md 2>/dev/null | sort -r | head -1
  2. If none, tell the user the chain does not exist and suggest list.
  3. Read the file fully and internalize it as working context.
  4. List the 5 most recent links:
    • ls .claude/chains/<chain-name>/*.md 2>/dev/null | sort -r | head -5
  5. Detect insufficiency and ask questions if needed:
    • Missing context referenced by the link
    • Ambiguous "Next Step"
    • Stale link (>7 days old)
  6. Show the user: chain name, most recent link filename, short summary, next step, and the 5 most recent links.
  7. If any insufficiency flags exist, stop and clarify before proceeding.

list

List all chains with counts and most recent link.

Steps:

  1. Check for chains:
    • ls -d .claude/chains/*/ 2>/dev/null
  2. For each chain, show:
    • chain name (directory)
    • link count (ls .claude/chains/<chain-name>/*.md 2>/dev/null | wc -l)
    • most recent link filename (ls .claude/chains/<chain-name>/*.md 2>/dev/null | sort -r | head -1)
  3. If no chains exist, suggest link to start one.