Borrowing it
Nothing to install: this file belongs to arcasilesgroup/ai-engineering. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/arcasilesgroup/ai-engineering/main/.agents/skills/ai-visual-recap/SKILL.mdgit clone --depth 1 https://github.com/arcasilesgroup/ai-engineeringWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/arcasilesgroup/ai-engineering/ai-visual-recap)<a href="https://agentmods.dev/skills/arcasilesgroup/ai-engineering/ai-visual-recap"><img src="https://agentmods.dev/badge/skills/arcasilesgroup/ai-engineering/ai-visual-recap.svg" alt="Measured on agentmods" height="20"></a>- NVIDIA SkillSpector pass
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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00147 | $0.00829 |
| Opus 5 | $0.00073 | $0.00415 |
| Sonnet 5 | $0.00029 | $0.00166 |
| Haiku 4.5 | $0.00015 | $0.00083 |
Grade A, and why
ai-visual-recap 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 8d 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.
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.
Recap the work unit as one page
What it produces
.ai/reports/NNN-recap-<slug>.html — a committed record of one range's change, printed in
chat as a file:// link beside the spec digest it carries.
Steps
-
Read
policy/visual-pages.mdfirst; it owns the budgets and the editorial bar. If that path is absent, refuse to author the page and restore the file — a recap written against remembered budgets is the drift the file exists to prevent. -
Scope the whole work unit, not the last fix: the implementation, the follow-ups, the tests and the record changes this thread made. Separate them from unrelated dirty files that existed before; if the boundary is genuinely unclear, say the assumption in one line and proceed.
-
Pick the base the range starts from — the approval commit for a build, the merge base for a PR — and run the command that derives the page from real bytes:
ai-eng report recap --spec <NNN> --base <ref> --summary "<1-3 paragraphs>"The file-tree and every excerpt come from
git diffover that range; the renderer refuses an excerpt that is not a real hunk. The summary is yours, and every claim in it names a file the diff touches. -
Show the link the command prints. A recap nobody clicks is a file.
-
Skip the page for a change that reviews faster as a plain diff — a small, single-file or obvious one. A recap is review overhead; say "one file, plain diff is enough" and stop.
-
UI changes need the visual named. When the range changed rendered UI, the summary names the components the diff touched and the before/after wireframes live on the plan's view page, where blocks are authored — the recap is derived from bytes and invents none.
-
Never re-read the whole diff twice: the budgets are the cost ceiling as much as the shape rule, and a recap that dumps is a recap the reviewer scrolls past.
Done when
ai-eng report recapexited 0, the page matches the reports-home shapedoctorchecks, and itsfile://link is in the reply.- The page's file list equals
git diff --name-statusover the same range.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 8d ago First seen · 67 lines · 147 tokens per session scan A 3af98a5878fd
ai-visual-recap is a skill published in the GitHub repository arcasilesgroup/ai-engineering (54 stars, last pushed 7d ago), licensed Apache-2.0. It adds 147 tokens to every session and 829 once invoked, about $0.0007 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 skills, from other repositories
github-navigator
GitHub operations via gh CLI. CRITICAL: Use instead of WebFetch for any github.com URL or GitHub repo path like owner/repo. Use when the user asks to inspect repositories, files, issues, pull requests, releases, Actions runs, or repository structure. Use when the user says 'show README', 'list issues', 'check PR'…
gpt-image-2-api
Generate and edit images with gpt-image-2 through a third-party OpenAI-compatible API using .env-configured OPENAIAPIKEY and OPENAIBASEURL values. Use when the user asks to create images, edit reference images, create visual assets, illustrations, or image variations with a configurable API endpoint.
pr-classify
A pull-request review process that classifies findings as Critical, Important, or Optional. A pull request is a proposed code change submitted for review before it is merged into the project.
pr-recheck
Second-pass re-review of a PR that already has review comments. Re-reads the diff and new commits against the existing open threads, marks each as addressed / partial / not addressed, resolves the addressed ones, and then either approves (only when everything is clean) or drafts new inline comments. Use when the user…
commit
A workflow for turning current code changes into small, logically grouped Git commits. A commit is a saved set of changes in Git that can be reviewed or reverted together.
pr-comment
A tool for drafting a GitHub pull request’s description from the current branch’s changes and the repository’s own template. It returns the result as raw Markdown that can be pasted into GitHub.