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/migsilva89/imark/imark-reviewgit clone --depth 1 https://github.com/migsilva89/imarkWhat 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 | $0.00016 | $0.00470 |
| Opus 5 | $0.00008 | $0.00235 |
| Sonnet 5 | $0.00003 | $0.00094 |
| Haiku 4.5 | $0.00002 | $0.00047 |
Grade A, and why
imark-review scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
Open a document in Imark for the user to review, and wait for their decision.
Run this, passing the arguments through as they came:
node "${CLAUDE_PLUGIN_ROOT}/scripts/imark.mjs" review $ARGUMENTS
Markdown only — plans, specs, RFCs, docs. If you are asked to review code, say that Imark is a markdown reader and that reviewing code is a job for GitHub or their editor.
The review happens on the file itself — no copy is made. The reviewer's
notes are written into the document as <!-- imark … --> blocks and stay
there; they travel with the file. Content piped on stdin, or several files at
once, gets a temporary stand-in document instead, cleaned up afterwards.
The command blocks until the user presses Approve or Send Back in the app. That is expected and can take a long time: do not interrupt it, do not put a timeout on it, and do not ask whether they are done.
When it returns, the output carries the decision and the notes, and on a refusal it also carries the steps to take. Follow them in the order they come — in particular, reply to the user before rewriting anything, and ask rather than guess when a note is ambiguous or two of them contradict each other.
Refer to notes by the words they are attached to — on “X” — rather than by number, because that is how the user sees them in the app.
When you act on a note, do not delete it. Mark it done by adding
resolved="<ISO date>" to its opening <!-- imark line — the note is the
reviewer's record of what was asked, the mark is yours of having done it. The
app shows resolved notes faded, and the next review round will not repeat them
back to you.
If the command says Imark is not installed, tell the user where the document ended up and carry on without the review.
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 · 44 lines · 16 tokens per session scan A 506f016829d1
imark-review is a command published in the GitHub repository migsilva89/imark (50 stars, last pushed 3d ago), licensed MIT. It adds 16 tokens to every session and 470 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.
Other commands, from other repositories
capture
Capture live screens/windows or ingest video; adaptive frames + contact sheet.
click
Target UI elements via peekaboo click.
see
Capture annotated UI maps with peekaboo see.
agent
Drive Peekaboo’s autonomous agent via peekaboo agent.
app
Control macOS apps via peekaboo app.
bridge
Diagnose Bridge hosts and verify signed operation receipt bundles.