refresh

A mid-workstream checkpoint command that saves pending knowledge and a resumable session record before the context window is reset. The context window is the amount of conversation an agent can keep available at once.

In plain words
What is it for?
Use it during unfinished work, such as when coding is complete but testing remains, to record the checkpoint, shut down teammates, and continue after context compaction and bootstrapping.
Why use it?
It reduces the risk of losing progress when a long task outgrows the current context, while leaving the work active rather than closing or merging it.

Command

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 commands/synaptic-labs-ai/pact-plugin/refresh
Clone the repo
git clone --depth 1 https://github.com/Synaptic-Labs-AI/PACT-Plugin
Per session 24 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,291 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 $0.00024 $0.02291
Opus 5 $0.00012 $0.01145
Sonnet 5 $0.00005 $0.00458
Haiku 4.5 $0.00002 $0.00229

Measured 2d ago against content hash a95d80a94466, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

refresh 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 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.

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.

pact-plugin/commands/refresh.md · 151 lines

How it starts

The opening of the file, as written. The whole thing — 151 lines — stays where its author put it; the contents beside it link to each section on GitHub.

PACT Refresh Protocol

Reset the context window mid-workstream without losing state. Refresh harvests pending knowledge, persists a resumable checkpoint to the session journal, shuts down teammates, and stands by for the user to run /compact (canonical) and then /PACT:bootstrap. The workstream CONTINUES after the reset — nothing is closed, merged, or cleaned up.

Cross-reference: For a PR-open pause with later resumption, see pause.md. For end-of-session cleanup and close, see wrap-up.md.


When to Use / When NOT

Command Purpose Workstream state after
/PACT:pause PR open, resume the review later Paused (PR open)
/PACT:wrap-up End-of-session cleanup + close Ended
/PACT:refresh Mid-workstream context checkpoint Active — continues after /compact + /PACT:bootstrap

Use refresh when the context window is near exhaustion while work is still mid-flight — for example CODE complete but TEST pending, a live HALT/algedonic signal, or worktrees holding unmerged commits. No orchestration retrospective runs (the arc is mid-flight, not ending); /PACT:pin-memory is optional; the harvest and its drain confirmation are MANDATORY.


Constraints (apply to every step)

  • LEAD-FRAME-ONLY: every step that spawns, stops, or messages teammates runs in the lead frame. The team roster is flat — teammates cannot spawn or stop teammates.
  • Never clean worktrees, branches, or commits. Verification is read-only; the workstream continues after the reset.
  • Never complete or delete tasks — the feature task, phase tasks, and any HALT/algedonic signal tasks must survive for resume.
  • Never write under {config_dir}/pact-refresh/. That directory is legacy state from a removed subsystem; checkpoint state lives ONLY in the session journal. {config_dir} is this session's Claude config root — the value of $CLAUDE_CONFIG_DIR when set and non-empty, otherwise $HOME/.claude. Read it off an absolute path the platform already injected into your context — your plugin root is {config_dir}/plugins/… — rather than shelling out for the variable. Substitute it before running any command; never assume ~/.claude.

Read the full file on GitHub · 151 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 · 151 lines · 24 tokens per session scan A a95d80a94466

Subscribe to this mod's changes

refresh is a command published in the GitHub repository Synaptic-Labs-AI/PACT-Plugin (71 stars, last pushed 2d ago), licensed MIT. It adds 24 tokens to every session and 2,291 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.