pause

A command for safely stopping the current session’s background coding work at a resumable point. It first allows near-complete pull requests to finish merging, then records what remains for later recovery.

In plain words
What is it for?
Use it when leaving a coding session, closing your laptop, or temporarily parking work while keeping eligible pull requests moving and recording unfinished tasks.
Why use it?
It prevents unfinished work from being left running when you need to close your laptop or stop for the day. It also preserves a machine-readable and human-readable resume point.

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/auerbachb/claude-code-config/pause
Any agent
npx skills add auerbachb/claude-code-config --skill pause
Clone the repo
git clone --depth 1 https://github.com/auerbachb/claude-code-config

Made for: Claude Code, Codex.

Per session 73 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 6,498 The whole file, excluding the scripts and references it only reads on demand.
Security scan B 1 finding. 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 $0.00073 $0.06498
Opus 5 $0.00036 $0.03249
Sonnet 5 $0.00015 $0.01300
Haiku 4.5 $0.00007 $0.00650

Measured 2d ago against content hash 205c1d1263f7, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade B, and why

pause scanned grade B with 1 finding 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 2d ago.

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.

Asks for rootmediumPrivilege escalation

A mod that escalates privileges can change anything on the machine, not only the project.

chmod 644 "$MARKER_PATH"
.claude/skills/pause/SKILL.md · 437 lines

How it starts

The opening of the file, as written. The whole thing — 437 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Land what can land. Park the rest at a deliberate boundary. Write a resume point you can pick up cold.

Two outputs, in this order: an active wind-down that drives near-done PRs to merged, then a machine-readable marker and human-readable summary the companion /pause-resume reads when you sit back down.

This command never relaxes a gate. cr-merge-gate.md Steps 1–1d, 1b and the Step 2 AC verification bind unchanged. The hard stops in CLAUDE.md "PR MERGE AUTHORIZATION" (human CHANGES_REQUESTED on HEAD, failing/incomplete CI, unresolved threads, unchecked AC, protection-modifying bypass) are hard stops here. The window never makes a borderline PR eligible; it only decides how long to wait on already-eligible ones.

One parameter: --window Nm. Fifteen minutes is the default graceful shutdown runway and triage threshold; a caller may choose a shorter or longer non-negative window. The command stops earlier when all owned work is terminal. At the chosen deadline it stops leftovers; it never merely returns while billable work continues.

Step 0: Resolve helpers and parse arguments

/pause is invocable from any thread — including one whose cwd is not this repo. Resolve each helper the same way every other stop-style command does:

resolve_script() {
  local name="$1" candidate
  for candidate in \
    "$HOME/.claude/skills-worktree/.claude/scripts/$name" \
    "$HOME/.claude/scripts/$name"; do
    if [[ -x "$candidate" ]]; then echo "$candidate"; return 0; fi
  done
  return 1
}
SESSION_STATE_SH=$(resolve_script session-state.sh)     || SESSION_STATE_SH=""
HANDOFF_STATE_SH=$(resolve_script handoff-state.sh)     || HANDOFF_STATE_SH=""
MERGE_GATE_SH=$(resolve_script merge-gate.sh)           || MERGE_GATE_SH=""
AC_CHECKBOXES_SH=$(resolve_script ac-checkboxes.sh)     || AC_CHECKBOXES_SH=""
PR_STATE_SH=$(resolve_script pr-state.sh)               || PR_STATE_SH=""
LOCAL_REVIEW_SH=$(resolve_script local-review.sh)       || LOCAL_REVIEW_SH=""
CLEAN_BEHIND_SH=$(resolve_script clean-behind-check.sh) || CLEAN_BEHIND_SH=""
EXECUTION_PAUSE_SH=$(resolve_script execution-pause.sh) || EXECUTION_PAUSE_SH=""
TASK_REGISTRY_SH=$(resolve_script background-task-registry.sh) || TASK_REGISTRY_SH=""

# Resolve repository identity independently of session-state persistence so a
# degraded state helper does not make the recovery marker undiscoverable.
REPO_KEY=""
if [[ -n "$SESSION_STATE_SH" ]]; then
  REPO_KEY=$("$SESSION_STATE_SH" --repo-key 2>/dev/null) || REPO_KEY=""
fi
if [[ -z "$REPO_KEY" ]]; then
  REMOTE=$(git remote get-url origin 2>/dev/null) || REMOTE=""
  REPO_KEY=$(printf '%s' "$REMOTE" | sed -e 's|\.git$||' \
    -e 's|.*github\.com[:/]\([^/]*/[^/]*\)$|\1|')
  [[ "$REPO_KEY" =~ ^[A-Za-z0-9._-]+/[A-Za-z0-9._-]+$ ]] || REPO_KEY=""
  REPO_KEY=$(printf '%s' "$REPO_KEY" | tr '[:upper:]' '[:lower:]')
fi

Resolve background-task-shutdown.md from the skills-worktree, then the home install location, and read it in full before Step 1. Never fall back to the current checkout for scripts or instruction documents: /pause can be invoked from an unrelated or untrusted repository, whose executable bit is not a trust signal.

An unresolved session-state.sh is degraded-but-continue — say so when reporting and carry on, because the marker still gets written. An unresolved merge-gate.sh means the landing phase cannot classify PRs reliably; skip landing for all PRs and park them, naming the missing helper in each parked entry's waiting_on field.

An unresolved TASK_REGISTRY_SH, unreadable shutdown reference, or durable registry-failure marker makes the shutdown audit unavailable. Arm every gate that can still be armed, skip any claim that no tasks are live, and carry the missing control into Step 8 as INCOMPLETE SHUTDOWN; never interpret an unreadable inventory as an empty one.

Parse --window Nm: extract the integer N from the first --window argument; default to 15. --window 0 means skip landing and checkpoint time, then stop everything immediately. Accept a non-negative integer with an optional trailing m; normalize leading zeroes as decimal, and reject non-numeric, negative, or greater-than-1440-minute values. Stop processing --window arguments after the first valid value. Compute T_end once:

Read the full file on GitHub · 437 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. 2d ago First seen · 437 lines · 73 tokens per session scan B 205c1d1263f7

Subscribe to this mod's changes

pause is a skill published in the GitHub repository auerbachb/claude-code-config (5 stars, last pushed 2d ago), licensed MIT. It adds 73 tokens to every session and 6,498 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it B with 1 finding (asks for root). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other skills, from other repositories

fixer

Surgical code fixer for Bug Hunter. Implements minimal, precise fixes for verified bugs. Uses doc-lookup (Context Hub + Context7) to verify correct API usage in patches. Respects fix strategy classifications (safe-autofix vs manual-review vs larger-refactor).

codexstar69/bug-hunter · 57 tokens

referee

Final arbiter for Bug Hunter. Receives Hunter findings and Skeptic challenges, independently re-reads code, and delivers authoritative verdicts with CVSS scoring and proof-of-concept generation for security findings.

codexstar69/bug-hunter · 44 tokens

commit-security-scan

Scan code changes for security vulnerabilities using Bug Hunter-native artifacts and STRIDE context. Use whenever the user asks for PR security review, commit-diff scanning, staged-change security checks, branch-comparison security review, or pre-merge security analysis of changed code.

codexstar69/bug-hunter · 57 tokens

decompose

Decompose project or track into modules with dependency mapping. Project scope updates architecture.md and derives .ai-context.md. Track scope generates hld.md (always) and lld.md (when --lld or High-complexity module triggers it) — design-mandated artifacts that drive implement, deploy-checklist, and upload sign-off.

drafthq/draft · 71 tokens

adr

Create and manage Architecture Decision Records. Documents significant technical decisions with context, alternatives, and consequences. Also supports evaluate (assess proposals) and design (system design) modes.

drafthq/draft · 37 tokens

deploy-checklist

Pre-deployment verification checklist. Generates customized checklists based on tech-stack with rollback triggers. Auto-invoked by /draft:upload.

drafthq/draft · 32 tokens