agent-fleet-manager skill
A skill and SQLite engine that watch thousands of sources on a schedule, detect changes by hash and keep an audit ledger.
- 107 ★ on the repository
- Last commit
Extended access
Why this level
- The fetcher visits network addresses given by the user and writes data to a local database.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/agent-fleet-managerDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The fetcher visits network addresses given by the user and writes data to a local database.
This is third-party code. Review the repository files before installing.
What it does
agent-fleet-manager consists of the `agentfleet` engine on the Python standard library and a `SKILL.md` that teaches an agent to run the fleet. Sources are registered with `add-source` and a check cadence. The `sweep` command queues what is due, leases batches of work to workers, collects results and compares content hashes. A worker can be a plain HTTP fetcher or the agent itself: it reads the manifest, checks the source and writes a result file into its own write scope. State lives in one SQLite file with five tables. Failures push the next check back with a growing interval, and `export-otel` emits OpenTelemetry traces.
Who it is for. People who track competitor, regulator or job pages at scale and want a verifiable change log.
Good fit when
- Competitor prices, changelogs and listings must be tracked on a schedule.
- Regulator pages are checked and the time of a change must be proven by a hash.
- Datasets and docs must stay current without reprocessing unchanged content.
- Several agents share one fleet and every write must be attributed.
Not a fit when
- The source cannot be fetched under site terms or robots.txt.
- A hosted monitoring system with a dashboard is needed.
- There are few sources and manual checks are enough.
Example request
Add the careers pages of five companies to the fleet with a daily check, run a sweep and show whose hash changed.Limitations
The engine only schedules, records and compares; the meaning of a route is defined by your worker. The built-in fetcher uses plain HTTP and does not log in to sites. The skill asks not to register sources you have no right to fetch and to respect robots.txt. TypeSafe classification is optional and needs their API key. The project is young: the repository has a small test set and one example. Sources behind login, dynamic loading or bot protection need a custom worker.
How to disable. Delete the skill folder and the `fleet.db` database file. If you put a periodic `sweep` into a scheduler, remove it separately.
FAQ
How do I add a source?
`python -m agentfleet.cli add-source --db fleet.db <name> <route> --cadence <seconds>`. The engine does not interpret the route.
What does sweep do?
It queues what is due, leases batches, receives results and compares hashes.
Are dependencies required?
No, only the Python standard library. Tests run with `python -m unittest discover -s tests -v`.
Can I watch it in Grafana?
Yes. `export-otel` posts traces over OTLP/HTTP to any collector.
Who can act as a worker?
The built-in HTTP fetcher, an external script or the agent itself. The contract is one: read the manifest, check the source and drop a result file into its own scope.
Discussion
No questions or reviews yet.
Related
SkillsAI Agent Skills by sanjay3290
AI Agent Skills
Cross-platform agent skills: read-only PostgreSQL, MySQL and MSSQL queries, Google Workspace, Jira, Telegram, text-to-speech and research
Skills for HR and recruiters: role requirements, sourcing on BOSS Zhipin and Liepin, resume review, interviews and a candidate ledger
Skills for coding agents that build and verify workspaces on the Lemma platform, with their agents and apps
A self-hosted control plane for running several agents: task dispatch, run review, spend tracking, connected via MCP, CLI or REST
Skill and CLI for job applications: fills Greenhouse, Ashby and Lever forms and logs only confirmed submissions
Sleepless Agent is a background-job CLI for Claude Code with a SQLite queue that retains pending work after a session ends.
README in short
The README sets out the five-table model, scheduling and retry rules, a quickstart, bring-your-own workers and OpenTelemetry export. A separate section covers TypeSafe classification.