Borrowing it
Nothing to install: this file belongs to agigante80/actual-mcp-server. 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/agigante80/actual-mcp-server/main/.claude/skills/working-overnight/SKILL.mdgit clone --depth 1 https://github.com/agigante80/actual-mcp-serverWrote 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/agigante80/actual-mcp-server/working-overnight)<a href="https://agentmods.dev/skills/agigante80/actual-mcp-server/working-overnight"><img src="https://agentmods.dev/badge/skills/agigante80/actual-mcp-server/working-overnight/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/skills/agigante80/actual-mcp-server/working-overnight"><img src="https://agentmods.dev/badge/skills/agigante80/actual-mcp-server/working-overnight.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00105 | $0.01030 |
| Opus 5 | $0.00053 | $0.00515 |
| Sonnet 5 | $0.00021 | $0.00206 |
| Haiku 4.5 | $0.00011 | $0.00103 |
Grade A, and why
working-overnight 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 12d 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 — 80 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Working overnight
Governed unattended work while the user is away. The engine is /loop: the user
launches /loop working-overnight and each cycle does one unit of work and returns
with fresh context. This skill holds the policy: what to pull, how to stay safe,
when to defer, and when to wind down. The overnight-continue hook keeps a single
cycle from halting on a question; it is not the engine.
Read references/safety.md (the tier model and the never-list) and
references/pipeline.md (the per-task governance pipeline) before the first cycle.
They are binding.
Two modes, told apart by the manifest
- Pre-flight (interactive, once, when the user sets up the night): arm the run.
- Cycle (each
/loopiteration, unattended): do one unit of work.
If .claude/overnight/active.md is absent, this is pre-flight. If it is present,
this is a cycle.
Pre-flight (arming)
Do this with the user present. Never start autonomous work until they confirm.
- Agree the work source and scope: which tickets, which investigations, and any no-go areas (paths, systems, ticket labels to avoid).
- Agree the budget (token or cost ceiling) and the investigation-depth cap (how many new tickets an investigation may create before implementation must catch up).
- Confirm the safety config in
references/safety.md: the allowed tools and the Tier-3 never-list. Do not proceed if the requested work would need anything on the never-list. - Ensure
.claude/overnight/is gitignored. - Write the manifest to
.claude/overnight/active.md(scope, budget, caps, no-go list, start time) and seed.claude/overnight/queue.mdwith the initial items. Creating the manifest arms the hook. - Tell the user to run
/loop working-overnightand leave. Stop here.
Cycle (one unit of work)
- Re-read
.claude/overnight/active.md,.claude/overnight/queue.md, and the project rules (CLAUDE.md, whose coding standards are inline). Fresh every cycle: this is why rule-drift does not accumulate. - Pick the next item by priority: a cleanly-gated ready ticket, else an ungated ticket to gate, else a bounded investigation. If no safe item remains, go to Wind-down.
- Classify the item by tier (
references/safety.md). A Tier-3 action is never performed: park it as needs-human and pick another item. - Resolve any question. Try web search, the code, and the project rules first.
Only a genuine you-only decision blocks: if it is low-stakes and reversible,
record the assumption for the report and continue; otherwise append it to
.claude/overnight/decisions.md(item, options, your recommendation) and pick another item. - Do the work through
references/pipeline.md. - Implementation lands in a per-item git worktree as a branch and a PR, never merged. Investigations write findings and create/gate tickets.
- Update
.claude/overnight/queue.md(done, parked, or newly created items). - Backstop: if context or the budget is near its limit, go to Wind-down now rather than starting another item.
What ships with it
2 files 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.
- 12d ago First seen · 80 lines · 105 tokens per session scan A 31362c1ac2b7
working-overnight is a skill published in the GitHub repository agigante80/actual-mcp-server (54 stars, last pushed yesterday), licensed MIT. It adds 105 tokens to every session and 1,030 once invoked, about $0.0005 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
add-matrix
Add Matrix channel integration via Chat SDK. Works with any Matrix homeserver.
add-resend
Add Resend (email) channel integration via Chat SDK.
add-macos-statusbar
Add a macOS menu bar status indicator for NanoClaw. Shows a bolt icon with a green/red dot indicating whether NanoClaw is running, with Start, Stop, and Restart controls. macOS only.
trend-analysis
A data-analysis workflow that groups records, rates performance, applies different growth assumptions, predicts values, and creates comparison charts.
excel-basic-statistics-and-routing
An Excel workflow for filtering grouped data, calculating averages, extracting row ranges, removing duplicates, and adding totals.
excel-conditional-filtering-optimization
A workflow for filtering Excel spreadsheet rows using several numeric conditions and exporting the matching results. Excel is a spreadsheet format that can contain multiple worksheets and large tables.