Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/AutomateLab-tech/agency-osnpx agentmods add commands/automatelab-tech/agency-os/refreshWrote 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/automatelab-tech/agency-os/refresh)<a href="https://agentmods.dev/commands/automatelab-tech/agency-os/refresh"><img src="https://agentmods.dev/badge/commands/automatelab-tech/agency-os/refresh/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/automatelab-tech/agency-os/refresh"><img src="https://agentmods.dev/badge/commands/automatelab-tech/agency-os/refresh.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 | $0.00000 | $0.00587 |
| Opus 5.5 | $0.00000 | $0.00235 |
| Sonnet 5.5 | $0.00000 | $0.00117 |
| Haiku 4.5 | $0.00000 | $0.00059 |
Grade A, and why
refresh 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 1mo 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
Command: refresh
Auto-enumerate the agent-runnable To-Do set and write it to state/todo-ids.json. No arguments. The operator's only job upstream is to mark rows in Notion with Exec=Agent; refresh then fetches them via the Notion REST API and the sidecar is the enumeration substrate for run.
The currently installed Notion MCP does not expose property-filtered enumeration of a data source, so this command shells out to scripts/query-tasks.py, which posts to POST /v1/data_sources/{id}/query with a server-side Status="To-Do" AND Exec="Agent" filter (Notion API version 2025-09-03). The integration token (NOTION_KEY in .env) must be shared with the Tasks database.
Run:
python .claude/skills/agency-os/scripts/query-tasks.py
The script:
- Loads
NOTION_KEYfrom.envandtasks_database.data_source_idfromreferences/notion-pointers.json. - Queries the data source with the two-gate filter, paginating through
has_more/next_cursor. - For each result, fetches the page's block children once to extract a
description_preview(the text between theDescriptionH2 and the next H2; first 200 chars). - Writes
.claude/skills/agency-os/state/todo-ids.json:{ "refreshed_at": "<iso>", "tasks": [ { "id": "<uuid>", "url": "https://www.notion.so/...", "title": "...", "corpus": "General", "priority": "3", "effort": "M", "type": "one-time", "cadence": null, "last_done": null, "exec": "Agent", "parent_task_id": null, "has_todo_subtasks": false, "description_preview": "<first 200 chars of Description>", "dependencies": [ { "id": "<uuid>", "status": "Done" }, { "id": "<uuid>", "status": "To-Do" } ] } ] } - Prints a summary:
refreshed: <N> agent-runnable To-Do tasks -> state/todo-ids.jsonfollowed by one line per task.
Failure modes. The script aborts with a non-zero exit and an explanatory message if NOTION_KEY is missing, the integration is not shared with the database, or the API returns an error. The existing sidecar is overwritten only after the query succeeds end-to-end.
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.
- 1mo ago First seen · 48 lines · 0 tokens per session scan A 998a878e023a
refresh is a command published in the GitHub repository AutomateLab-tech/agency-os (3 stars, last pushed 4mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 587 tokens. 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
p9
A Tech Lead mode for writing prompts and managing the P8 team. A Tech Lead guides technical work and breaks projects into tasks.
architecture-note
Record a brief Architecture Decision Record (ADR) about a recent technical choice.
write-handoff
Generate a dense state-restoration document before ending a session.
dashboard-session-fork
Fork a session into a new one. Usage /dashboard:session-fork.
dashboard-session-hide
Hide a session from the dashboard list. Usage /dashboard:session-hide.
sdd
A command that creates three connected planning files for a software feature: requirements, design, and tasks.