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 HazelKahlil/Kahlil-skills --skill multica-squad-buildergit clone --depth 1 https://github.com/HazelKahlil/Kahlil-skillsWrote 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/hazelkahlil/kahlil-skills/multica-squad-builder)<a href="https://agentmods.dev/skills/hazelkahlil/kahlil-skills/multica-squad-builder"><img src="https://agentmods.dev/badge/skills/hazelkahlil/kahlil-skills/multica-squad-builder/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/hazelkahlil/kahlil-skills/multica-squad-builder"><img src="https://agentmods.dev/badge/skills/hazelkahlil/kahlil-skills/multica-squad-builder.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 | $0.00054 | $0.00950 |
| Opus 5 | $0.00027 | $0.00475 |
| Sonnet 5 | $0.00011 | $0.00190 |
| Haiku 4.5 | $0.00005 | $0.00095 |
Grade A, and why
multica-squad-builder 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 — 72 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Multica Squad Builder
Purpose
Output a tailored setup pack for one project, not a fixed team. Frontier models need few procedural constraints; keep hard gates and platform facts. Every generated rule must earn its place with evidence.
Team model
- Leader: real leader. Writes specs, scales effort, packs briefs, judges output, unblocks with decisions.
- Workers / Builders: research, implement, test, open PRs in an assigned packet.
- Reviewer: optional, risk-triggered only. Analyst: optional, spec/evidence support on request.
- Human Owner: final authority for product, merge, deploy, and irreversible calls.
Choose topology first: direct single agent (no squad) when one executor and one deliverable suffice; routed single-worker when the issue is squad-assigned but one Worker owns it; staged or parallel squad only for real dependencies or independent work packages.
Leader Charter
- Spec before dispatch. On intake, write goal, acceptance criteria, and out-of-scope. For unclear work, ask ONE sharp question or decide and record the decision. Medium/large work does not start without a spec.
- Scale effort. Work smaller than the cost of delegating (small edits, direct answers, status cleanup) the leader does personally. Spawn children only for parallel, bulky, or isolatable work.
- Self-contained briefs. Every child issue states objective, expected output shape, boundaries, and decisions already made (stack, style, edge-case policy) so members inherit decisions.
- Judge, don't forward. On every wake, check member output against the spec — accept, reject with specifics, or replan. Reports to the owner are the leader's synthesized conclusion.
- Unblock with a decision. When a member is stuck, pick a path with a one-line rationale. Escalate to the Human Owner only for irreversible, permission, or direction calls, always with a recommendation and a default.
- Mechanical duties. Verify dispatch receipts: a new task in
queued/dispatched/running; an attempt is not a dispatch. Check existing children before creating new ones. Report only read-back facts.
What ships with it
16 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- agents/openai.yaml 294 B
- CHANGELOG.md 8.5 KB
- README.md 1.0 KB
- references/leader-playbook.md 2.8 KB
- references/multica-primitives.md 2.8 KB
- references/role-theme-presets.md 1.5 KB
- references/safety-gates.md 2.5 KB
- references/squad-design.md 2.4 KB
- scripts/validate_skill.py 3.1 KB runs code
- templates/agent-analyst-instructions.md 518 B
- templates/agent-leader-instructions.md 2.0 KB
- templates/agent-reviewer-instructions.md 721 B
- templates/agent-worker-instructions.md 1.2 KB
- templates/setup-pack-template.md 1.6 KB
- templates/squad-instructions.md 1.7 KB
- VERSION 6 B
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 · 72 lines · 54 tokens per session scan A 9b1eaa4230ba
multica-squad-builder is a skill published in the GitHub repository HazelKahlil/Kahlil-skills (5 stars, last pushed 19d ago), licensed MIT. It adds 54 tokens to every session and 950 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
yao-geo-tracking
Build a company-specific GEO backend tracking plan from a company name plus optional supporting information, using authoritative retrieval anchored on the official website, business-feature recognition, and market-aware GEO monitoring methods, then deliver a direct-vs-indirect attribution design, data model, roadmap…
yao-geo-comparison-builder
A tool for creating evidence-based Chinese brand comparison content for AI search and answer platforms. It compares a target brand with competitors, traditional approaches, or self-built solutions.
yao-geo-explainer-builder
A tool for creating detailed, evidence-based educational articles about a topic for Chinese AI search and answer platforms. It can include explanations, how-to steps, selection guidance, common mistakes, FAQs, and a glossary.
yao-geo-page-audit
Diagnose a website page or small page set for GEO readiness with authoritative public evidence, systematic page analysis, code/content/schema fixes, and four-format Chinese report delivery.
yao-geo-brand-graph
A method for building a brand entity knowledge graph from company information, websites, products, customers, people, places, evidence sources, and use cases. A knowledge graph records entities and the supported relationships between them.
yao-deepseek-crawler
Use when a user provides DeepSeek web AI-search keywords, repeat count, target entity, and entity type, then needs repeated fresh-window crawls aggregated into JSON plus a Kami HTML GEO report. Not for generic website crawling, DeepSeek API chat, SEO writing, or one-off answer generation.