Borrowing it
Nothing to install: this file belongs to Eduardo-Salvador/Agent-Harness-Kit. 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/Eduardo-Salvador/Agent-Harness-Kit/main/.agents/skills/parallel-dispatch/SKILL.mdgit clone --depth 1 https://github.com/Eduardo-Salvador/Agent-Harness-KitWrote 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/eduardo-salvador/agent-harness-kit/parallel-dispatch)<a href="https://agentmods.dev/skills/eduardo-salvador/agent-harness-kit/parallel-dispatch"><img src="https://agentmods.dev/badge/skills/eduardo-salvador/agent-harness-kit/parallel-dispatch.svg" alt="Measured on agentmods" 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 | $0.00055 | $0.00447 |
| Opus 5 | $0.00028 | $0.00224 |
| Sonnet 5 | $0.00011 | $0.00089 |
| Haiku 4.5 | $0.00006 | $0.00045 |
Grade A, and why
parallel-dispatch 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 yesterday.
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
Parallel dispatch
Follow ../../../harness/playbooks/parallel-execution.md. Activate automatically after graph reconciliation whenever at least two nodes are ready; the user does not need to name this skill.
- Read the host-evidenced numeric implementation capacity and the current graph revision. If parallel contexts are unavailable, record
sequential-fallback; do not claim parallel execution. - Run
agent-harness schedule harness-state/TASK-GRAPH.md --capacity <host-capacity>(or the equivalent packaged CLI) and use only its selected IDs. The scheduler rechecks completed dependencies, accepted assurance, active capacity, and normalized write collisions. - For the whole selected batch, pin task specs, reserve distinct leases/isolations/model-dispatch routes under the expected graph revision, and create
harness.parallel-dispatch/v1asreserving. Never dispatch from conversational memory. - Invoke the actual Codex subagent/task operation once for every selected node without waiting for earlier children to complete. Use separate calls/references, explicit model/reasoning overrides, and the smallest declared context packet. Calls may return sequentially; the children must remain running concurrently.
- Persist each returned agent/task reference and adapter evidence, transition only confirmed children to
active, and report the active worker count. - Wait for the first child completion or attention event and refill free capacity. At 60–90 seconds without observable progress, warn/request a checkpoint; on the second consecutive occurrence, interrupt, preserve state, and reassign or serialize within budget.
- Dependents become ready only after all declared dependencies and assurance gates pass. Cross-area results meet in an explicit integration node; never let parallel children edit shared integration/generated outputs concurrently.
Automatic fan-out, refill, and fan-in are normal execution and require no repeated human approval. Human input is requested only for a genuine authority/product/risk decision.
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.
- yesterday Changed bd5416df886e
- 6d ago First seen · 19 lines · 55 tokens per session scan A becd0b9c6e4f
parallel-dispatch is a skill published in the GitHub repository Eduardo-Salvador/Agent-Harness-Kit (6 stars, last pushed 2d ago), licensed MIT. It adds 55 tokens to every session and 447 once invoked, about $0.0003 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 2026-08-31.
Other skills, from other repositories
qa
Run scalable, isolated live QA for nac development. The top-level local orchestrator must parse n (default 4), dispatch one setup worker with this skill, copy its n assignment contracts verbatim into exactly n parallel test workers with this skill, then dispatch one aggregate worker with this skill using all test…
release
Cut and publish a full stable NAC release after main, release-PR, and publication CI pass. Use when a maintainer asks for a stable version bump, tag, or GitHub Release. Never use for release candidates; NAC RC releases are automated.
openrig-user
Use when a specific rig command, subcommand, or flag is already known and you need its exact syntax, JSON shape, defaults, or error meaning. NOT for natural capability discovery, open-ended how-do-I questions, or choosing which OpenRig move applies.
triage
Triage a GitHub repository's open issues by finding exact duplicates, rejecting evidenceably off-base requests, requesting concrete clarification, applying only existing labels, and opening a linked root-cause issue when multiple reports share one underlying invariant failure. Use when a maintainer asks to triage…
orienting-to-an-inherited-seat
Use when you have just been primed into an EXISTING seat through a planned handover — a different agent retired and handed you the seat plus its earned context — and you need a world model of what just happened to you. Covers how a handover differs from compaction and from a fresh launch, the…
forming-an-openrig-mental-model
Use when the system around you does not make sense yet: you just booted into a seat and do not know how the pieces fit; someone said rig, pod, seat, fleet, topology, or slice and you are not certain what they mean here; you are unsure what kind of rig you are in or what it is for; you do not know how skills reach you…