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.
npx agentmods add commands/devzonayed/mochi/checkpointgit clone --depth 1 https://github.com/DevZonayed/MochiWhat 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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00017 | $0.00888 |
| Opus 5 | $0.00009 | $0.00444 |
| Sonnet 5 | $0.00003 | $0.00178 |
| Haiku 4.5 | $0.00002 | $0.00089 |
Grade A, and why
checkpoint scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 67 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are writing a new link in the continuum context chain for this project. A "link" is a small, durable summary of what happened since the previous link — it survives session boundaries and compaction.
What to do
-
Read the previous state to know what to summarize against:
.continuum/STATE.md— current project truth (do not duplicate; only note deltas).- The last entry in
.continuum/chain/index.jsonl— tells you the previous link id and timestamp. - The previous link's
summary.md(highest-numbered dir under.continuum/chain/links/) — so you don't repeat decisions already captured. - Frontend verification status (only if applicable):
If this reports any UNVERIFIED frontend files or FAILURES, include them as Open threads in the new link summary so they survive into the next session. Don't silently skip — surfacing unfinished verification is the whole point of the loop.node "$(cat .continuum/.plugin-root)/lib/verify_status_cli.js"
-
Compose a link summary in your head with these sections (omit any that are empty). Hard cap: ~800 tokens. Be terse; pointers, not payloads.
## Decisions - <new settled decisions; if any SUPERSEDES a prior decision, say so explicitly with the prior link id> ## Changes since last link - <what changed at the level of intent — not line-by-line. Reference git commits, not diffs.> ## Open threads - <unresolved questions in flight as of this checkpoint> ## Constraints / Do not - <new "do not" items; e.g. "client rejected X — do not re-propose"> ## Refs - <semantic anchors → where to find the detail. e.g. "auth-decision → commit abc123, file src/auth.ts:L40-80">Rules:
- NEVER restate code that git already holds. Use commit ids.
- NEVER narrate conversation flow ("then the user asked"). Record outcomes only.
- If something contradicts a previous link, STATE the contradiction and which is authoritative now — do not silently overwrite.
-
Write the link by piping your summary to the helper, which handles ids, timestamps, git commit capture, and index update atomically:
cat <<'CONTINUUM_SUMMARY_EOF' | node "$(cat .continuum/.plugin-root)/lib/write_link.js" --tags "tag1,tag2" <YOUR SUMMARY MARKDOWN HERE> CONTINUUM_SUMMARY_EOFThe script prints the new link id (e.g.
0003). Choose tags that aid retrieval later: domain (auth,db,ui), kind (decision,bugfix,refactor,bootstrap), or topic. -
Regenerate
STATE.mdto reflect new cumulative truth:- Read the OLD
.continuum/STATE.md. - Apply the deltas from this new link: superseded decisions are REMOVED (not crossed out), new decisions added, open threads updated, "do not" list updated.
- Hard cap: 150 lines. If you can't fit, suggest a rollup (do not silently truncate).
- Use
Writeto overwrite.continuum/STATE.md.
- Read the OLD
-
Report back to the user in 2 lines: new link id, and one-sentence headline of what it captures.
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.
- yesterday First seen · 67 lines · 17 tokens per session scan A 20eafb309db5
checkpoint is a command published in the GitHub repository DevZonayed/Mochi (3 stars, last pushed 15d ago), licensed MIT. It adds 17 tokens to every session and 888 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 2026-08-31.
Other commands, from other repositories
toon
A lightweight data format designed to minimize token usage when passing structured data to LLMs.
browse
You have access to a browser via the MCP Browser Bridge tools. Use them to interact with web pages.
computer
Control the macOS desktop — move mouse, click, type, press keys, scroll, drag, take screenshots, list windows/displays, clipboard. Uses usecomputer CLI for native Quartz event automation.
autoresearch
Autonomous research loop — iteratively edit, test, measure, keep/discard. Usage: /autoresearch [--budget 5m] [--provider ark].
gitlab-scrum
GitLab Scrum management - create/manage issues, labels, milestones, and boards. Use when user says: create issue, list issues, move issue, add label, scrum, kanban, backlog, or task tracking.
web-setup
Guide users through Chrome extension setup for browser control. Use when user asks about setup, installing the extension, connecting Chrome, or when browsercontrol tools return 'No browser connected'.