fleet-eval
Skill1 starsChanged 50 days agoMITView on GitHub
Verify a claim against reality — shapes assertions into {subject, expected, probe, evidence, verdict} and runs independent probes. Use when another agent reports X works/ships/is deployed and proof is needed before trusting it, or when the user says verify this claim or fleet-eval.
In this repository113 other files it gives its agents.
- AGENTS.mdAGENTS.md
- ac-ideateSkill
- ac-listSkill
- ac-roadmapSkill
- agentbridge-mcpSkill
- almanacSkill
- almquist-shellSkill
- ansible-fleetSkill
- apfsSkill
- apple-siliconSkill
- arena-actorsSkill
- asustor-nasSkill
- a-teamSkill
- atem-miniSkill
- at-tuiSkill
- autoresearchSkill
- babysitSkill
- bonzibuddySkill
- bop-brainstormSkill
- bop-catalogSkill
- bop-onSkill
- bop-teamSkill
- bop-workflowSkill
- bsd-portabilitySkill
- code-refactoringSkill
- codex-personasSkill
- context-graphSkill
- councilSkill
- datadog-backupSkill
- dd-tagsSkill
- delegateSkill
- deming-pdsaSkill
- deterministic-replaySkill
- dispatch-nuSkill
- durable-commitSkill
- elixir-gastown-opsSkill
- ergo-ircSkill
- esp32-tang-bridgeSkill
- fleet-ircSkill
- freebsd-build-vmSkill
- freebsd-piSkill
- freebsdSkill
- ganttSkill
- gastown-opsSkill
- gastown-slingSkill
- gh-authSkill
- gitea-releaseSkill
- gitea-runnerSkill
- henshuSkill
- irclaw-cluster-opsSkill
- irclaw-research-teamSkill
- irc-taskSkill
- jail-factorySkill
- jjSkill
- jobcard-systemSkill
- kubernetesSkill
- leptosSkill
- loreSkill
- lower-bound-reviewSkill
- macvlan-netSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 10
Every claim of form "X is deployed / working / reachable / shipped" must pass:
- Distinctive: line 80
When another agent says "X is deployed / working / shipped", invoke:
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
fleet-eval — Claim -> Proof Protocol
The gate
Every claim of form "X is deployed / working / reachable / shipped" must pass:
- A deterministic probe (command or API call)
- Return expected evidence (string, status code, file hash, etc.)
- Emit a structured verdict JSON
Claim types + standard probes
| Claim shape | Probe | Expected evidence |
| "service X at URL is up" | curl -sS -w '\nHTTP %{http_code}\n' URL --max-time 10 | HTTP 2xx |
| "file F exists with content C" | sha256sum F | expected SHA |
| "pkg P installed on host H" | ansible H -m raw -a "pkg info P" | pkg info output |
| "process P running on host H" | ansible H -m raw -a "pgrep -af P" | non-empty |
| "nick N in IRC channel" | (use fleet-irc skill with NAMES query) | nick present |
| "Gitea release R exists" | tea release view --repo ... --tag R | release JSON |
| "schema file valid JSON Schema" | python -c "import jsonschema; jsonschema.Draft202012Validator.check_schema(json.load(open('F')))" | no exception |
| "VM V alive" | ssh ... echo ALIVE OR pgrep -af qemu.*V | "ALIVE" or non-empty |
Output contract (always)
{
"claim": "string — the original claim being verified",
"probe": "string — exact command run",
"expected": "string — what constituted pass",
"evidence": "string — actual probe output, truncated to 1KB",
"verdict": "PASS | FAIL | INCONCLUSIVE",
"ts": "ISO8601",
"took_ms": <number>
}Written to stdout as JSON; also appended to /tmp/fleet-eval.log.
Usage patterns
Invoke after every claim
fleet-eval verify "Phase 16 VM is ready" \
--probe "ssh -J studio@10.0.3.1 -p 2225 builder@localhost echo ALIVE" \
--expect "ALIVE"Batch verify (multi-claim reports)
fleet-eval batch < claims.jsonl
# Each line: {"claim": "...", "probe": "...", "expected": "..."}Negative expected (expected-fail)
fleet-eval verify "Integrations install fails missing datadog-agent" \
--probe "..." --expect-exit 1 --expect-stderr-match "missing dependency"When NOT to use fleet-eval
- Pure planning / brainstorming — nothing to verify yet
- Questions about what something IS (use other skills)
Composes with
- Any agent making deployment/infra claims
superpowers:verification-before-completion(complementary; this is the runtime version)datadog-metrics(emit eval results as metrics)
Related
/Users/studio/rust-town/fleet-ops/scripts/fleet-eval.nu— the runnable/Users/studio/rust-town/fleet-ops/scripts/drift-monitor.nu— hourly drift checker/Users/studio/rust-town/fleet-ops/data/declared-alive.toml— drift inventory
Reflex pattern (for any invoking agent)
When another agent says "X is deployed / working / shipped", invoke:
nu /Users/studio/rust-town/fleet-ops/scripts/fleet-eval.nu verify "CLAIM" --probe "CMD" --expect "VALUE"One probe, one verdict. No retry loops. Proof > report.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.