Getting it into your agent
There is no command for this one: it runs only inside a plugin, and the catalogue could not identify which plugin ships it. The source is linked below.
Wrote 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/jeroromano/claude-code-orchestration-template/delegation-protocol)<a href="https://agentmods.dev/skills/jeroromano/claude-code-orchestration-template/delegation-protocol"><img src="https://agentmods.dev/badge/skills/jeroromano/claude-code-orchestration-template/delegation-protocol.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 | $0.00089 | $0.06173 |
| Opus 5 | $0.00044 | $0.03086 |
| Sonnet 5 | $0.00018 | $0.01235 |
| Haiku 4.5 | $0.00009 | $0.00617 |
Grade B, and why
delegation-protocol scanned grade B with 1 finding 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 6d 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.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
| Pre-merge review of a Claude-authored branch (every behavioral branch) | `/codex:review --base <base> --background` - no per-invocation model/effort flags: it inherits Codex's RESOLVED config, which is the repo-local ` How it starts
The opening of the file, as written. The whole thing — 78 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Delegation protocol
Delegation is only cheaper than main-thread work if (a) the noisy work happens in someone else's context and (b) the task spec is smaller than the work itself. This protocol enforces both.
1. Task spec template (mandatory for any writing delegation)
GOAL: <one sentence, observable outcome>
ALLOWED FILES: <explicit paths or globs - the writer may touch nothing else>
NON-GOALS: <what must NOT change; include "no refactors" by default>
ACCEPTANCE: <verifiable criteria, bullet list>
VALIDATION: <exact commands: test / lint / build - from CLAUDE.md project commands>
RISK PATH: <yes/no, per CLAUDE.md risk paths. If yes: adversarial review is mandatory>
EXPECTED DIFF SHAPE: <files x approximate size - lets any reviewer spot drift instantly>
If VALIDATION cannot be filled, the project commands in CLAUDE.md are missing: ask the human to fill them before delegating. A spec without validation commands is decorative.
2. Routing table
| Work | Route | Why |
|---|---|---|
| Mechanical, fully specified | fast-worker agent | cheapest capable writer, isolated context |
| Scoped implementation with an approved spec | Codex rescue, background: /codex:rescue --fresh --background --model gpt-5.6-sol --effort medium <approved spec>; raise to --effort high only for hard bugs or multi-module work. Sol implements specs; it never authors them. Sol is a limited-access preview: if the CLI rejects the model, rerun without --model (CLI default) or fall back to fast-worker (§5) |
separate quota pool + cross-family perspective |
| One hard scoped question, no writing | deep-reasoner agent | premium reasoning without premium main thread |
| Escalation: deep-reasoner insufficient, or one decision's stakes justify premium spend | premium-reasoner agent - requires human authorization; include PREMIUM-APPROVED in the prompt | top-tier one-shot reasoning, spend-gated; premium tiers draw shared limits at a multiple of default models - verify with /usage |
| Pre-merge review of a Claude-authored branch (every behavioral branch) | /codex:review --base <base> --background - no per-invocation model/effort flags: it inherits Codex's RESOLVED config, which is the repo-local .codex/config.toml only when that file exists and the repo is trusted, and otherwise the global ~/.codex/config.toml (§4). Effort: the highest the path exposes (§5); on the direct path the raw CLI can additionally pin per invocation - codex review --commit <sha> -c model_reasoning_effort="max" (verified August 2026) - a SINGLE-COMMIT form: it reviews that commit only, never everything a merge would introduce, so it does not substitute for a branch-level review; for a branch, pin max in the resolved config and use the flagless command. Delivered via the review transport (§6): on native Windows auto ships this same review as an inline-task instead - the direct command hangs there until the per-machine sandbox fix (§6) |
cross-family, read-only, cheap on quota |
| Pre-merge review of a Codex-authored branch | diff-reviewer agent or a human - never Codex | no agent approves its own authorship |
| Risk-path change review, Claude-authored branch | /codex:adversarial-review --base <base> plus a focus instruction from CLAUDE.md risk paths, at the same MAXIMUM effort as every other review (§5) - there is no higher rung to escalate to: max is the top of the enum, and the ceiling is set by the invocation path (the companion's --effort flag stops at xhigh; the TOML and the raw CLI accept max). Same transport rule (§6): under inline-task this pass ships inline with the focus text in the prompt and no --effort pin unless preflight (§5) shows the resolved config below the ceiling |
decorrelated failure modes where bugs are most expensive |
| Risk-path change review, Codex-authored branch | diff-reviewer agent (or a human) with the CLAUDE.md risk-path focus text - never Codex | the adversarial pass obeys the same rule: no agent approves its own authorship |
| Codex exhausted, absent, code that must stay local, or a Codex-authored diff | diff-reviewer agent | keeps the second pass; on Claude-authored diffs it will note the weaker same-family decorrelation |
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.
- 6d ago First seen · 78 lines · 89 tokens per session scan B cc2d22b1862e
delegation-protocol is a skill published in the GitHub repository jeroromano/claude-code-orchestration-template (1 stars, last pushed 28d ago), licensed MIT. It adds 89 tokens to every session and 6,173 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
ospec
Document-driven OSpec workflow for initialization, change/goal routing, validation, archiving, and durable project knowledge.
plan-change
Turns a change idea into a stable spec.md and actionable plan.md under Docs/ /Specs/ /. Opens with a fixed prompt for goals/context, then conversationally resolves scope and naming; proceeds through workflow steps in order. Tone: concise, cheerful, inquisitive. Does not create the Specs folder, spec.md, or plan.md…
inferring-a-spell-from-examples
Use when the user invokes /capture-this-chat or /build-spell --from-transcript. Reads a transcript and produces a context dictionary plus a draft SKILL.md, which hands off to building-a-spell at Stage 2.
building-a-subagent-spell
Use when the meta-builder routes to kind=subagent. Generates a spell that dispatches one or more subagents to do work in parallel or in isolation.
intuitive-interviewing
Use when conducting any interview-style flow with a user (building a spell, refining one, gathering requirements). Picks depth, filters questions by relevance, matches against known shapes, and detects mid-stream pivots.
building-a-workflow-spell
Use when the meta-builder routes to kind=workflow. Generates a workflow spell with explicit stages, checkpoints, and loop-back conditions.