Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add Knuckles-Team/repository-manager --skill repository-manager-parallel-lane-orchestrationgit clone --depth 1 https://github.com/Knuckles-Team/repository-managerWrote 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/knuckles-team/repository-manager/repository-manager-parallel-lane-orchestration)<a href="https://agentmods.dev/skills/knuckles-team/repository-manager/repository-manager-parallel-lane-orchestration"><img src="https://agentmods.dev/badge/skills/knuckles-team/repository-manager/repository-manager-parallel-lane-orchestration/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/skills/knuckles-team/repository-manager/repository-manager-parallel-lane-orchestration"><img src="https://agentmods.dev/badge/skills/knuckles-team/repository-manager/repository-manager-parallel-lane-orchestration.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 | $0.00191 | $0.01912 |
| Opus 5 | $0.00096 | $0.00956 |
| Sonnet 5 | $0.00038 | $0.00382 |
| Haiku 4.5 | $0.00019 | $0.00191 |
Grade A, and why
repository-manager-parallel-lane-orchestration 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 11d 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.
How it starts
The opening of the file, as written. The whole thing — 139 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Parallel lane orchestration
One orchestrator, many workers. Workers build; the orchestrator decomposes, dispatches, adjudicates, lands, and prunes. The orchestrator does not do the building — its scarce resource is judgment, not throughput.
The loop
- Decompose into lanes partitioned by file, not by ticket id. Reserved id blocks stop id collisions; they do not stop two lanes implementing the same fix twice in the same file. If two lanes would touch one file, sequence them instead.
- Dispatch each lane with the full briefing contract below. Route mechanical and well-scoped work to a cheaper model; keep the orchestrator on the expensive one.
- Land as they arrive. The moment a lane reports, merge its branch into
mainlocally, then delete the branch and remove its worktree. Do not batch. A queue of unlanded branches becomes a conflict pile, and worktrees left behind are the single largest source of workspace drift. - Push at the end, once the program is coherent — not per lane.
Lane briefing contract
Every dispatch states all of these, every time. Omitting one reliably produces a lane that stalls, edits a shared checkout, or reports unverified work as done.
- Isolation: take a real
git worktree add <root>/<repo>/<branch> -b <branch> main. Never edit the canonical checkout — a background sync resets it and discards the work. - Never use a harness worktree-isolation flag against a shared multi-worktree repo. It
writes
core.bare = trueinto the shared config and every linked worktree of that repo starts failinggit status/git commitinvisibly. - Do not push. Commit locally; leave the branch for the orchestrator.
- Never
--no-verify. A scopedSKIP=<hook-ids>is allowed only for whole-repo, non-differential gates failing on pre-existing state, and only with evidence in the commit message that the diff touches none of the implicated paths. - Do not park on a background watcher. Waiting on a notification that never fires is the most common way a lane silently burns its whole budget. Re-run the command to check state.
- Test invocation, verbatim, including the project-specific runner. A bare test command that resolves the system interpreter produces confident false verdicts at scale.
- Commit identity. Use the repo's configured identity. Never commit as a name that a privacy/identity gate derives its banned-token list from at runtime.
- Report shape: separate what was verified live from what was inferred from reading code, and say plainly when something is written but unverified.
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.
- 11d ago First seen · 139 lines · 191 tokens per session scan A ebd0fd9d1684
repository-manager-parallel-lane-orchestration is a skill published in the GitHub repository Knuckles-Team/repository-manager (2 stars, last pushed 14d ago), licensed MIT. It adds 191 tokens to every session and 1,912 once invoked, about $0.0010 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
coding-agents-farm
To orchestrate parallel coding-agent farms (Claude, Codex, Copilot, Gemini, etc.) on isolated git worktrees.
fable-thinking
Reasoning protocol distilled from Claude Fable 5.1. Makes any model reason like Fable — evidence-grounded claims, multi-hypothesis diagnosis, concrete simulation, adversarial self-review, calibrated outcome-first delivery. Its never-skipped Floor check catches simple-looking trick questions models answer confidently…
agent-framework-py-release
Use when cutting a Python release for the microsoft/agent-framework monorepo. Triggers on "bump py versions", "cut a python release", "prepare release PR for python", "release py packages", "bump python to X.Y.Z", or similar requests to bump Python package versions and prepare a release PR. Handles all four lifecycle…
atmos-hooks
Atmos hooks: lifecycle events, hook kinds, command/store/git/security hooks, step/steps hooks, when: conditions, scoping and overrides, toolchain integration, --skip-hooks, and Atmos Pro/local output.
atmos-pro
Atmos Pro setup and workflows: settings.pro, GitHub OIDC, affected and inventory uploads, stack locks, pro commit, workflow dispatch, merge queues, and drift detection.
atmos-git
Atmos Git and GitOps: git.repositories, clone/pull/status/diff/commit/push/clean, local Git hook shims, signed commits, managed workdirs, fork-PR trust gate, and auth via identities or github/sts.