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 agentmods add skills/partikularwaters/banka/delegatenpx skills add partikularwaters/Banka --skill delegategit clone --depth 1 https://github.com/partikularwaters/BankaWhat 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 | $0.00064 | $0.03879 |
| Opus 5 | $0.00032 | $0.01939 |
| Sonnet 5 | $0.00013 | $0.00776 |
| Haiku 4.5 | $0.00006 | $0.00388 |
Grade A, and why
delegate 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 2d 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 — 349 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Delegation only pays off if each ticket can be executed correctly by a fresh session that has never seen this conversation and has no access to the reasoning that produced the plan. Model capability and session separation are different: a Junior-safe ticket may run on the same, a lighter, or a stronger model, while Senior-required work must run in the user-selected senior-capability mode.
Context Contract
Required: an approved implementation plan from charter — Step 0 stops
without one · the resolved architecture file, to check Absolute Invariants
and security/compliance boundaries when classifying each item's tier.
Conditional: the existing delegation queue, when it already has unstarted or in-progress tickets — read before appending, never silently overwritten.
Excluded by default: anything outside the approved plan's scope — this skill splits an already-approved plan, it does not re-plan.
Outputs: the written queue (three tier-grouped lists, full ticket specs for all three, and a unified Execution Sequence showing every ticket's dependencies across tiers), one ready-to-paste handoff block per Junior-safe ticket in dependency order, and a plain checklist per Owner-required item — never a session handoff, since no AI session executes those.
Write authority: the tier-resolved delegation-queue.md only — append
tickets with stable, never-reused numbers; never touch the code the tickets
describe.
Fresh-session isolation covers conversation context, not working files. Banka's safe default is serial execution in one checkout: finish and survey one ticket before another session edits the same working directory. Parallel ticket execution is allowed only when each session receives a separate Git worktree and branch, with its results reviewed and merged deliberately. A local or cloud-hosted model may execute a ticket if its host can read the assigned project state, edit the assigned checkout/worktree, and run the required verification; Banka does not launch models or create worktrees itself.
This skill does not write code. It writes tickets. Writing a bad ticket is more expensive than writing no ticket — a bad ticket costs a full fresh-session context load, produces wrong output, and then needs the dredge skill to fix. Take the time to make each one right.
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.
- 2d ago First seen · 349 lines · 64 tokens per session scan A 9a4581ecf066
delegate is a skill published in the GitHub repository partikularwaters/Banka (3 stars, last pushed 3d ago), licensed MIT. It adds 64 tokens to every session and 3,879 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
agent-harness-fault-injection
Use when an agent workflow needs deterministic recovery evidence for sandbox, MCP/tool, worker, checkpoint, memory, or orchestration failures.
agent-framework-azure-ai-py
Build persistent agents on Azure AI Foundry using the Microsoft Agent Framework Python SDK.
agent-orchestration-improve-agent
Systematic improvement of existing agents through performance analysis, prompt engineering, and continuous iteration.
agent-orchestrator
Meta-skill que orquestra todos os agentes do ecossistema. Scan automatico de skills, match por capacidades, coordenacao de workflows multi-skill e registry management.
agent-evaluation-reporting
Use when summarizing agent evaluations where autonomous, assisted, failed, timed-out, or invalid outcomes must remain distinct and comparable.
agent-orchestration-multi-agent-optimize
Optimize multi-agent systems with coordinated profiling, workload distribution, and cost-aware orchestration. Use when improving agent performance, throughput, or reliability.