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.
git clone --depth 1 https://github.com/yuri-semenenko/ai-engineering-workspaceWrote 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/commands/yuri-semenenko/ai-engineering-workspace/handoff)<a href="https://agentmods.dev/commands/yuri-semenenko/ai-engineering-workspace/handoff"><img src="https://agentmods.dev/badge/commands/yuri-semenenko/ai-engineering-workspace/handoff/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/yuri-semenenko/ai-engineering-workspace/handoff"><img src="https://agentmods.dev/badge/commands/yuri-semenenko/ai-engineering-workspace/handoff.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00024 | $0.00991 |
| Opus 5.5 | $0.00010 | $0.00396 |
| Sonnet 5.5 | $0.00005 | $0.00198 |
| Haiku 4.5 | $0.00002 | $0.00099 |
Grade A, and why
handoff 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 5d 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.
What it actually says
Handoff Prompt
Role: Act as a senior engineer handing a work stream to a fresh session, which may be your own next chat.
Context: A new chat starts with none of this conversation. The handoff carries what it needs to continue: the next step, the state, the decisions and their reasons. Anything the next session can re-read from git, the PR, or the ticket stays out.
Task: Write the handoff for the current work stream. If the user instead gives an existing handoff, resume from it: compare it with the live state (branch, latest commit, PR status), say what drifted, and state the next step in one line.
Constraints:
- Do not write files outside this workspace. Return the handoff as one Markdown block for the user to save outside the repository, for example at
~/.handoffs/<key>/<stream>.md, where<stream>is the ticket key (PROJ-123) or a kebab-case topic.<key>is the repository's directory name plus a short hash of itsoriginURL (or of the main checkout's path when there is noorigin), so two repositories with the same name do not share a folder. Give the user the command that prints it, for their shell, exactly as written: the hash covers exact bytes, and a hand translation prints a different key.- POSIX shell:
main=$(git worktree list --porcelain | sed -n '1s/^worktree //p'); id=$(git config --get remote.origin.url); [ -n "$id" ] || id=$main; printf '%s-%s\n' "$(basename "$main" .git)" "$(printf '%s' "$id" | git hash-object --stdin | cut -c1-8)" - PowerShell:
$enc = [Console]::OutputEncoding; [Console]::OutputEncoding = [Text.UTF8Encoding]::new($false); $main = (git worktree list --porcelain)[0] -creplace '^worktree '; $id = git config --get remote.origin.url; if (-not $id) { $id = $main }; [Console]::OutputEncoding = $enc; $f = [IO.Path]::GetTempFileName(); [IO.File]::WriteAllText($f, $id); $h = git hash-object $f; Remove-Item $f; '{0}-{1}' -f ((Split-Path -Leaf $main) -creplace '(?<=.)\.git$'), $h.Substring(0, 8)
- POSIX shell:
- Never suggest saving it inside the repository; it carries work details.
- One handoff per stream, replaced each time. It is state, not a log. Stay under about 60 lines and skip empty sections. Tell the user, before they save over an existing file at that path, to check that its repo line names this checkout's path and that its branch, PR, or ticket is this stream's. A matching branch alone is not enough: two clones of one repository share the key and often the branch. If only the repo path differs, replace the file only if this checkout moved; otherwise pick a different
<stream>. - A handoff saved before the key existed stays at
~/.handoffs/<repo>/<stream>.md; resume it by that path. When the new handoff replaces one at a different path, list the old path under Pointers asSupersedes <path>so it is deleted when the stream closes. - Authorizations given in this chat (commit, push, PR) are history, not permission. The next session asks again.
- A lesson that outlives the stream belongs in durable notes, not the handoff.
Output Format:
- Title line
# <stream>: <one-line goal>, then one line withRepoand the main checkout path, branch and short sha, PR link and state, ticket link - Next step: the first concrete action, specific enough to run without asking
- State: done / in progress / not started, one line each, with evidence
- Decisions: what was decided, by whom, and why, plus rejected options
- Constraints: rules the user set that bear on the next step
- Open: questions and what is being waited on
- Pointers: paths and links
- One line the user can paste into the next chat to resume
Success Criteria:
- A fresh session can take the next step without asking anything already answered here.
- Nothing in it is re-derivable from git, the PR, or the ticket.
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.
- 5d ago First seen · 40 lines · 24 tokens per session scan A 9d38568bbfc9
handoff is a command published in the GitHub repository yuri-semenenko/ai-engineering-workspace (4 stars, last pushed 3d ago), licensed MIT. It adds 24 tokens to every session and 991 once invoked, about $0.0001 per session on Opus 5.5. 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 2026-11-01.
handoff compared with commands from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| handoffyuri-semenenko/ai-engineering-workspace | Command | 4 | 3d ago | MIT | A | 24 | GitHub Copilot |
| al-context-createjaviarmesto/ALDC-AL-Development-Collection | Command | 115 | yesterday | MIT | A | 35 | Claude Code |
| al-memory-createjaviarmesto/ALDC-AL-Development-Collection | Command | 115 | yesterday | MIT | A | 36 | Claude Code |
| sdd-cleanGregorBiswanger/featherspec | Command | 70 | 1mo ago | MIT | A | 21 | Claude Code |
| sdd-setupGregorBiswanger/featherspec | Command | 70 | 1mo ago | MIT | A | 27 | Claude Code |
| 7_gofer_saveeai-support/eai-gofer | Command | 1 | 5d ago | Apache-2.0 | A | 22 | Claude Code |
| save-from-chatandiupn/andy-universal-agent-rules | Command | 2 | 4mo ago | MIT | A | 0 | Claude Code, Cursor |