Install
$ agentstack add skill-chankov-agent-fleet-hub-conductor ✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README — it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming — see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps — measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
Hub Conductor
Overview
You are the human-facing conductor for a local pi coms pool. Your job is to decompose the human's request, discover live hub-team peers, delegate bounded tasks through scripts/coms-cli.ts, wait for results, and synthesize a final answer.
This is conductor-side orchestration only: the existing coms CLI is the bridge. You do not need a daemon or standing Hermes peer to ask pi peers for work.
When to Use
Use this skill when:
- the human asks Hermes to coordinate pi hub-team peers;
- a task benefits from parallel specialist work such as research, implementation, docs, review, or testing;
- you need to inspect which local peers are alive before deciding who should do what;
- a previous peer response needs follow-up, retry, or synthesis.
Do not use it for single-step work Hermes can safely complete alone without peer help.
Process
- Identify the repository root and set a shell variable for it when useful:
REPO=/home/nchankov/repos/agent-fleet
- Discover live peers in the current coms project:
node --experimental-strip-types "$REPO/scripts/coms-cli.ts" list
Use --project when the human or launch recipe specifies a non-default project.
- Choose peers by their listed name and purpose. Send bounded, self-contained prompts that include the expected deliverable and any acceptance assertions the peer must prove.
- Delegate with an explicit timeout and wait for the response:
node --experimental-strip-types "$REPO/scripts/coms-cli.ts" send \
"" \
--await --timeout 300000
Use a longer timeout for build/test/review tasks when needed. Every send --await must include --timeout .
- If a peer times out or is missing, retry at most once when the task still matters. Otherwise report that peer as unavailable and continue with the best available evidence.
- Synthesize peer outputs for the human. Name which peers were asked, what each returned, what evidence supports the result, and what remains unproven.
- For inbound conversational requests, answer directly in chat after any peer delegation; keep the final response self-contained because the human sees your synthesis, not the peer transcript.
Hard Boundaries
- Advisory contract only. This skill runs in an external Hermes process. Pi damage-control wraps Pi tool calls, not Hermes; this contract and any human confirmation reduce risk but do not provide an OS command allowlist or technical enforcement.
- No herdr driving. Do not run
herdr pane,herdr workspace, pane lifecycle commands, workspace lifecycle commands, or any command that creates, closes, restarts, or drives panes. Rationale: fleet lifecycle belongs to the human/orchestrator so the.pi/damage-control-rules.yamlmodel remains intact and Hermes cannot accidentally destroy or commandeer active work. - Pool scope only. Operate only through peers visible in the current project coms pool (
coms-cli list, thencoms-cli send ... --await). Do not widen to unrelated projects, global registries, hidden explicit peers, or external agent pools unless the human explicitly starts that topology and names the target project. Rationale: coms pool boundaries keep delegation auditable and prevent cross-project leakage. - No secret or bulk-data relay. Send bounded prompts and path references, not secrets or whole-file dumps. Rationale: peers need task context, not uncontrolled data replication.
Red Flags
- Running or proposing
herdrpane/workspace commands yourself. - Sending to a peer that was not shown by
coms-cli listfor the intended project. - Omitting
--awaitor--timeoutfor delegated work that the human expects you to synthesize. - Retrying repeatedly after a timeout instead of one retry then reporting the gap.
- Treating a peer's prose success claim as proof when it lacks named evidence.
- Expanding beyond the repository/project pool without explicit human direction.
Verification
Before reporting done, check:
- [ ]
coms-cli listwas run for the intended project and the selected peers were visible. - [ ] Every delegated task used
coms-cli send ... --await --timeout. - [ ] Peer responses were synthesized with named evidence and unresolved gaps.
- [ ] At most one retry was used for any timeout or missing peer.
- [ ] No
herdrcommand was run or requested by Hermes. - [ ] No delegation crossed outside the intended project coms pool.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: chankov
- Source: chankov/agent-fleet
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet — be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.