suspend

A procedure for pausing all active development work with durable snapshots and handoff records.

In plain words
What is it for?
It helps snapshot worktrees, create draft handoffs, and record suspended work in the relevant issue files.
Why use it?
It preserves each agent's branch, changes, and task context so the work can be resumed after the current session ends.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/loopdive/js2/suspend
Any agent
npx skills add loopdive/js2 --skill suspend
Clone the repo
git clone --depth 1 https://github.com/loopdive/js2

Made for: Claude Code, Codex.

Per session 20 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,395 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00020 $0.01395
Opus 5 $0.00010 $0.00698
Sonnet 5 $0.00004 $0.00279
Haiku 4.5 $0.00002 $0.00139

Measured yesterday against content hash 18d12a15d413, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

suspend scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/skills/suspend/SKILL.md · 110 lines

How it starts

The opening of the file, as written. The whole thing — 110 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/suspend — suspend the standing team with durable handoffs

Everything in-flight becomes: a pushed branch + a draft PR whose description is the handoff + a ## Suspended Work section in the lane's issue file. A resumer needs NOTHING from this session's memory — worktrees, transcripts, and timers are all treated as already lost.

Run the phases in order. Phase 1 is time-critical (agents may be mid-write); everything after is durable bookkeeping.

Phase 1 — freeze and snapshot (do this FIRST, before any messaging)

  1. ListAgents — inventory running agents. For each, note its worktree path (/home/user/js2wasm/.claude/worktrees/agent-<id> or from its spawn record) and its ISSUE id.
  2. For each agent worktree — snapshot before talking to anyone:
    cd <worktree> && git add -A && \
    git -c user.name="Thomas Tränkler" -c [email protected] \
      commit --no-verify -m "wip(#<issue>): suspension snapshot ✓"
    
    • --no-verify is sanctioned for WIP snapshots (CI runs the real gates on the eventual real PR).
    • The commit lands on the worktree's CHECKED-OUT branch (verify with git branch --show-current), NOT necessarily the harness's worktree-agent-<id> ref — record the actual branch name; a resumer given the wrong ref finds plain main (this burned a lane on 2026-08-16, see #4484's record).
    • An empty snapshot (nothing to commit) is fine — note "clean tree at ".
  3. Only now message each agent (SendMessage): "SUSPEND — stop work; your worktree was snapshot at on ; do not commit further." Do not wait for acknowledgments; killed or busy agents are already covered by the snapshot.

Phase 2 — handoff into the issue files

For each lane (running agents AND merged-but-unshipped lanes), append to plan/issues/<id>-<slug>.md:

## Suspended Work (<UTC timestamp>)

- **Branch**: <branch> at <sha> (pushed to origin)
- **Worktree at suspension**: <path> (treat as gone; the branch is the truth)
- **State**: <one of: implementation complete + verification pending /
  mid-implementation, last verified step X / snapshot unverified — may not
  typecheck>
- **Verified so far** (runs the AGENT executed, with numbers): <...>
- **NOT yet verified / next steps in order**: <...>
- **Traps for the resumer**: <base predates waves X/Y — do not rebase, the
  lead reconciles at merge; tier-sensitive pins need
  `build-runtime-eval-provider.mjs --refusal-only`; etc.>

Read the full file on GitHub · 110 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 110 lines · 20 tokens per session scan A 18d12a15d413

Subscribe to this mod's changes

suspend is a skill published in the GitHub repository loopdive/js2 (59 stars, last pushed yesterday), licensed Apache-2.0. It adds 20 tokens to every session and 1,395 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.